hand and face again

This commit is contained in:
2026-01-18 14:43:14 -06:00
parent 716428ec0b
commit 9256050292
3 changed files with 551 additions and 647 deletions

3
.gitignore vendored
View File

@@ -1,4 +1,5 @@
asl_kaggle/
hand_landmarker.task
asl-dataset.zip
asl-signs.zip
asl-signs.zip
best_asl_transformer.pth

618
test.py
View File

@@ -1,170 +1,208 @@
import mediapipe as mp
import cv2
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import math
from collections import deque, Counter
import pandas as pd # ← added for rebuilding labels
# Modern MediaPipe Tasks API (no legacy solutions module)
import mediapipe as mp
from mediapipe.tasks import python
from mediapipe.tasks.python import vision
# PyTorch ≥ 2.6 checkpoint loading fix
import numpy as np
import numpy.core.multiarray
import numpy.dtypes
torch.serialization.add_safe_globals([
np.ndarray,
np.dtype,
np.dtypes.Int64DType,
np.core.multiarray._reconstruct
])
# Positional Encoding
# ===============================
# MODEL DEFINITION
# ===============================
class PositionalEncoding(nn.Module):
def __init__(self, d_model, max_len=100):
super(PositionalEncoding, self).__init__()
def __init__(self, d_model, max_len=128):
super().__init__()
pe = torch.zeros(max_len, d_model)
position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
pe[:, 0::2] = torch.sin(position * div_term)
pe[:, 1::2] = torch.cos(position * div_term)
pe = pe.unsqueeze(0)
self.register_buffer('pe', pe)
self.register_buffer('pe', pe.unsqueeze(0))
def forward(self, x):
return x + self.pe[:, :x.size(1), :]
return x + self.pe[:, :x.size(1)]
# Model architecture
class TransformerCNN_ASL(nn.Module):
def __init__(self, input_dim=77, num_classes=250, d_model=512, nhead=8, num_layers=6, dim_feedforward=2048):
super(TransformerCNN_ASL, self).__init__()
class TransformerASL(nn.Module):
def __init__(self, input_dim, num_classes, d_model=256, nhead=8, num_layers=4):
super().__init__()
self.proj = nn.Linear(input_dim, d_model)
self.norm_in = nn.LayerNorm(d_model)
self.pos = PositionalEncoding(d_model, max_len=128)
self.input_dim = input_dim
self.d_model = d_model
# Input projection
self.input_projection = nn.Linear(input_dim, d_model)
self.input_norm = nn.LayerNorm(d_model)
# Positional encoding
self.pos_encoder = PositionalEncoding(d_model, max_len=100)
# Transformer Encoder with Self-Attention
encoder_layer = nn.TransformerEncoderLayer(
enc_layer = nn.TransformerEncoderLayer(
d_model=d_model,
nhead=nhead,
dim_feedforward=dim_feedforward,
dropout=0.1,
dim_feedforward=d_model * 4,
dropout=0.15,
activation='gelu',
batch_first=True,
norm_first=True
)
self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
self.encoder = nn.TransformerEncoder(enc_layer, num_layers=num_layers)
# CNN Blocks for pattern detection
self.conv1 = nn.Conv1d(d_model, 1024, kernel_size=3, padding=1)
self.bn1 = nn.BatchNorm1d(1024)
self.pool1 = nn.MaxPool1d(2)
self.dropout1 = nn.Dropout(0.3)
self.head = nn.Sequential(
nn.LayerNorm(d_model),
nn.Dropout(0.25),
nn.Linear(d_model, num_classes)
)
self.conv2 = nn.Conv1d(1024, 2048, kernel_size=3, padding=1)
self.bn2 = nn.BatchNorm1d(2048)
self.pool2 = nn.MaxPool1d(2)
self.dropout2 = nn.Dropout(0.3)
self.conv3 = nn.Conv1d(2048, 4096, kernel_size=3, padding=1)
self.bn3 = nn.BatchNorm1d(4096)
self.pool3 = nn.AdaptiveMaxPool1d(1) # Global pooling
self.dropout3 = nn.Dropout(0.4)
# Fully connected layers
self.fc1 = nn.Linear(4096, 4096)
self.bn_fc1 = nn.BatchNorm1d(4096)
self.dropout_fc1 = nn.Dropout(0.5)
self.fc2 = nn.Linear(4096, 2048)
self.bn_fc2 = nn.BatchNorm1d(2048)
self.dropout_fc2 = nn.Dropout(0.4)
self.fc3 = nn.Linear(2048, 1024)
self.bn_fc3 = nn.BatchNorm1d(1024)
self.dropout_fc3 = nn.Dropout(0.3)
self.fc4 = nn.Linear(1024, num_classes)
def forward(self, x):
batch_size = x.size(0)
# Project to d_model
x = self.input_projection(x)
x = self.input_norm(x)
x = x.unsqueeze(1)
# Add positional encoding
x = self.pos_encoder(x)
# Transformer encoder with self-attention
x = self.transformer_encoder(x)
# Reshape for CNN
x = x.permute(0, 2, 1)
# CNN pattern detection
x = F.gelu(self.bn1(self.conv1(x)))
x = self.pool1(x)
x = self.dropout1(x)
x = F.gelu(self.bn2(self.conv2(x)))
x = self.pool2(x)
x = self.dropout2(x)
x = F.gelu(self.bn3(self.conv3(x)))
x = self.pool3(x)
x = self.dropout3(x)
# Flatten
x = x.view(batch_size, -1)
# Fully connected layers
x = F.gelu(self.bn_fc1(self.fc1(x)))
x = self.dropout_fc1(x)
x = F.gelu(self.bn_fc2(self.fc2(x)))
x = self.dropout_fc2(x)
x = F.gelu(self.bn_fc3(self.fc3(x)))
x = self.dropout_fc3(x)
x = self.fc4(x)
return x
def forward(self, x, key_padding_mask=None):
x = self.proj(x)
x = self.norm_in(x)
x = self.pos(x)
x = self.encoder(x, src_key_padding_mask=key_padding_mask)
x = x.mean(dim=1)
return self.head(x)
# Load the trained model
print("Loading model...")
checkpoint = torch.load('asl_kaggle_transformer.pth', map_location='cpu')
label_encoder = checkpoint['label_encoder']
num_classes = checkpoint['num_classes']
input_dim = checkpoint['input_dim']
config = checkpoint['model_config']
# ===============================
# FEATURE EXTRACTION
# ===============================
def get_features_sequence(landmarks_seq, max_frames=100):
if landmarks_seq is None or len(landmarks_seq) == 0:
return None, None
model = TransformerCNN_ASL(
input_dim=input_dim,
num_classes=num_classes,
d_model=config['d_model'],
nhead=config['nhead'],
num_layers=config['num_layers'],
dim_feedforward=config['dim_feedforward']
wrist = landmarks_seq[:, 0:1, :]
landmarks_seq = landmarks_seq - wrist
scale = np.linalg.norm(landmarks_seq[:, 9], axis=1, keepdims=True)
scale = np.maximum(scale, 1e-6)
landmarks_seq = landmarks_seq / scale[:, :, np.newaxis]
landmarks_seq = np.nan_to_num(landmarks_seq, nan=0.0, posinf=0.0, neginf=0.0)
landmarks_seq = np.clip(landmarks_seq, -10, 10)
tips = [4, 8, 12, 16, 20]
bases = [1, 5, 9, 13, 17]
curls = [np.linalg.norm(landmarks_seq[:, t] - landmarks_seq[:, b], axis=1)
for b, t in zip(bases, tips)]
curl_features = np.stack(curls, axis=1)
deltas = np.zeros_like(landmarks_seq)
if len(landmarks_seq) > 1:
deltas[1:] = landmarks_seq[1:] - landmarks_seq[:-1]
pos_flat = landmarks_seq.reshape(len(landmarks_seq), -1)
delta_flat = deltas.reshape(len(landmarks_seq), -1)
seq = np.concatenate([pos_flat, delta_flat, curl_features], axis=1)
T, F = seq.shape
if T < max_frames:
pad = np.zeros((max_frames - T, F), dtype=np.float32)
seq_padded = np.concatenate([seq, pad], axis=0)
else:
seq_padded = seq[:max_frames]
mask = np.zeros(max_frames, dtype=bool)
mask[:min(T, max_frames)] = True
return seq_padded.astype(np.float32), mask
# ===============================
# MANUAL DRAWING FUNCTION
# ===============================
HAND_CONNECTIONS = [
(0, 1), (1, 2), (2, 3), (3, 4),
(0, 5), (5, 6), (6, 7), (7, 8),
(0, 9), (9, 10), (10, 11), (11, 12),
(0, 13), (13, 14), (14, 15), (15, 16),
(0, 17), (17, 18), (18, 19), (19, 20),
(5, 9), (9, 13), (13, 17)
]
def draw_hand_landmarks(image, landmarks_list):
h, w = image.shape[:2]
# Draw connections (blue lines)
for start_idx, end_idx in HAND_CONNECTIONS:
start = landmarks_list[start_idx]
end = landmarks_list[end_idx]
start_pt = (int(start.x * w), int(start.y * h))
end_pt = (int(end.x * w), int(end.y * h))
cv2.line(image, start_pt, end_pt, (255, 0, 0), 2)
# Draw landmarks (green circles)
for lm in landmarks_list:
x = int(lm.x * w)
y = int(lm.y * h)
cv2.circle(image, (x, y), 5, (0, 255, 0), -1)
# ===============================
# MAIN PROGRAM
# ===============================
print("Loading trained model...")
checkpoint = torch.load("best_asl_transformer.pth", map_location="cpu")
model = TransformerASL(
input_dim=checkpoint['input_dim'],
num_classes=checkpoint['num_classes'],
d_model=checkpoint['d_model'],
nhead=checkpoint['nhead'],
num_layers=checkpoint['num_layers']
)
model.load_state_dict(checkpoint['model_state_dict'])
model.load_state_dict(checkpoint['model'])
model.eval()
total_params = sum(p.numel() for p in model.parameters())
print(f"Loaded Transformer+CNN model")
print(f"Total parameters: {total_params:,}")
print(f"Number of ASL signs: {num_classes}")
print(f"Sample signs: {label_encoder.classes_[:10]}")
# ─── FIX: Rebuild real sign names from train.csv ─────────────────────
print("\n" + "=" * 70)
print("Rebuilding sign name mapping from train.csv...")
# Setup MediaPipe
BaseOptions = mp.tasks.BaseOptions
HandLandmarker = mp.tasks.vision.HandLandmarker
HandLandmarkerOptions = mp.tasks.vision.HandLandmarkerOptions
VisionRunningMode = mp.tasks.vision.RunningMode
try:
# CHANGE THIS PATH to where your train.csv actually is
train_df = pd.read_csv("asl_kaggle/train.csv") # ← most important line!
# Get unique signs, sorted (same order LabelEncoder usually uses)
real_signs = sorted(train_df['sign'].unique())
# Use real sign names instead of numbers
label_encoder_classes = real_signs
print("SUCCESS! Loaded real sign names")
print("Number of classes:", len(real_signs))
print("First 15 signs:", real_signs[:15])
print("=" * 70 + "\n")
except Exception as e:
print("ERROR loading train.csv:", e)
print("Falling back to numeric labels (you'll see numbers instead of words)")
label_encoder_classes = checkpoint['label_encoder_classes']
print("First 15 (still numbers):", label_encoder_classes[:15])
print("=" * 70 + "\n")
# MediaPipe Tasks setup
BaseOptions = python.BaseOptions
HandLandmarker = vision.HandLandmarker
HandLandmarkerOptions = vision.HandLandmarkerOptions
VisionRunningMode = vision.RunningMode
MODEL_PATH = "hand_landmarker.task" # Make sure this file is in the folder
options = HandLandmarkerOptions(
base_options=BaseOptions(model_asset_path='hand_landmarker.task'),
base_options=BaseOptions(model_asset_path=MODEL_PATH),
running_mode=VisionRunningMode.VIDEO,
num_hands=1,
min_hand_detection_confidence=0.5,
@@ -174,270 +212,118 @@ options = HandLandmarkerOptions(
landmarker = HandLandmarker.create_from_options(options)
# Buffers
MAX_FRAMES = 100
sequence_buffer = []
prediction_buffer = deque(maxlen=15)
def get_optimized_features(hand_landmarks):
"""
Extract optimally normalized relative coordinates from MediaPipe landmarks
Returns 77 features
"""
# Extract raw coordinates
points = np.array([[lm.x, lm.y, lm.z] for lm in hand_landmarks], dtype=np.float32)
# Step 1: Translation invariance - center on wrist
wrist = points[0].copy()
points_centered = points - wrist
# Step 2: Scale invariance - normalize by palm size
palm_size = np.linalg.norm(points[9] - points[0]) # wrist to middle finger base
if palm_size < 1e-6:
palm_size = 1.0
points_normalized = points_centered / palm_size
# Step 3: Standardization
mean = np.mean(points_normalized, axis=0)
std = np.std(points_normalized, axis=0) + 1e-8
points_standardized = (points_normalized - mean) / std
# Flatten base features (63 features)
features = points_standardized.flatten()
# Step 4: Derived features
finger_tips = [4, 8, 12, 16, 20] # Thumb, Index, Middle, Ring, Pinky
# Distances between consecutive fingertips (4 distances)
tip_distances = []
for i in range(len(finger_tips) - 1):
dist = np.linalg.norm(points_normalized[finger_tips[i]] - points_normalized[finger_tips[i + 1]])
tip_distances.append(dist)
# Distance of each fingertip from palm center (5 distances)
palm_center = np.mean(points_normalized[[0, 5, 9, 13, 17]], axis=0)
tip_to_palm = []
for tip in finger_tips:
dist = np.linalg.norm(points_normalized[tip] - palm_center)
tip_to_palm.append(dist)
# Finger curl indicators (5 curls)
finger_curls = []
finger_bases = [1, 5, 9, 13, 17]
for base, tip in zip(finger_bases, finger_tips):
curl = np.linalg.norm(points_normalized[tip] - points_normalized[base])
finger_curls.append(curl)
# Combine all features: 63 + 4 + 5 + 5 = 77
all_features = np.concatenate([
features,
tip_distances,
tip_to_palm,
finger_curls
])
return all_features.astype(np.float32)
# Initialize webcam
cap = cv2.VideoCapture(0)
if not cap.isOpened():
print("Error: Cannot open webcam")
print("Cannot open webcam")
exit()
# Set camera resolution for better performance
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1280)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 720)
frame_count = 0
fps_counter = 0
fps_start_time = cv2.getTickCount()
current_fps = 0
# Prediction smoothing buffer
from collections import deque
prediction_buffer = deque(maxlen=10)
print("\n" + "=" * 60)
print("ASL Recognition - Transformer+CNN Model")
print("=" * 60)
print("Controls:")
print(" ESC - Exit")
print(" SPACE - Clear prediction buffer")
print(" 'h' - Toggle hand landmarks visibility")
print("=" * 60 + "\n")
print("\nASL Recognition running - Press ESC to quit")
print("Controls: ESC = quit | SPACE = clear | H = toggle landmarks\n")
show_landmarks = True
frame_timestamp_ms = 0
with torch.no_grad():
while True:
success, image = cap.read()
if not success:
print("Failed to read frame from webcam")
break
while cap.isOpened():
success, image = cap.read()
if not success:
break
# Flip image horizontally for mirror view
image = cv2.flip(image, 1)
image = cv2.flip(image, 1)
h, w = image.shape[:2]
# Convert to MediaPipe format
mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=image)
mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=image)
frame_timestamp_ms += 33
# Detect hands
results = landmarker.detect_for_video(mp_image, frame_count)
frame_count += 1
results = landmarker.detect_for_video(mp_image, frame_timestamp_ms)
# Calculate FPS
fps_counter += 1
if fps_counter >= 30:
fps_end_time = cv2.getTickCount()
time_diff = (fps_end_time - fps_start_time) / cv2.getTickFrequency()
current_fps = fps_counter / time_diff
fps_counter = 0
fps_start_time = cv2.getTickCount()
overlay = image.copy()
cv2.rectangle(overlay, (10, 10), (520, 340), (0, 0, 0), -1)
cv2.addWeighted(overlay, 0.65, image, 0.35, 0, image)
# Process hand landmarks if detected
if results.hand_landmarks and len(results.hand_landmarks) > 0:
hand_landmarks = results.hand_landmarks[0]
if results.hand_landmarks:
hand_landmarks_list = results.hand_landmarks[0]
# Draw hand landmarks if enabled
if show_landmarks:
# Draw connections
connections = [
(0, 1), (1, 2), (2, 3), (3, 4), # Thumb
(0, 5), (5, 6), (6, 7), (7, 8), # Index
(0, 9), (9, 10), (10, 11), (11, 12), # Middle
(0, 13), (13, 14), (14, 15), (15, 16), # Ring
(0, 17), (17, 18), (18, 19), (19, 20), # Pinky
(5, 9), (9, 13), (13, 17) # Palm
]
if show_landmarks:
draw_hand_landmarks(image, hand_landmarks_list)
# Get image dimensions
h, w = image.shape[:2]
current_frame = np.array(
[[lm.x, lm.y, lm.z] for lm in hand_landmarks_list],
dtype=np.float32
)
# Draw connections
for connection in connections:
start_idx, end_idx = connection
start = hand_landmarks[start_idx]
end = hand_landmarks[end_idx]
sequence_buffer.append(current_frame)
if len(sequence_buffer) > MAX_FRAMES:
sequence_buffer = sequence_buffer[-MAX_FRAMES:]
start_point = (int(start.x * w), int(start.y * h))
end_point = (int(end.x * w), int(end.y * h))
if len(sequence_buffer) >= 10:
seq_np = np.array(sequence_buffer)
feats, mask = get_features_sequence(seq_np, MAX_FRAMES)
cv2.line(image, start_point, end_point, (0, 255, 0), 2)
if feats is not None:
x = torch.from_numpy(feats).float().unsqueeze(0)
key_padding_mask = torch.from_numpy(~mask).unsqueeze(0)
# Draw landmarks
for i, landmark in enumerate(hand_landmarks):
x = int(landmark.x * w)
y = int(landmark.y * h)
with torch.no_grad():
logits = model(x, key_padding_mask=key_padding_mask)
probs = F.softmax(logits, dim=-1)[0]
pred_idx = torch.argmax(probs).item()
conf = probs[pred_idx].item()
# Different colors for different parts
if i == 0: # Wrist
color = (255, 0, 0)
radius = 8
elif i in [4, 8, 12, 16, 20]: # Fingertips
color = (0, 0, 255)
radius = 6
else:
color = (0, 255, 0)
radius = 4
# Now using real sign names!
sign = label_encoder_classes[pred_idx]
cv2.circle(image, (x, y), radius, color, -1)
cv2.circle(image, (x, y), radius + 2, (255, 255, 255), 1)
if conf > 0.40:
prediction_buffer.append(sign)
# Extract features
features = get_optimized_features(hand_landmarks)
final_sign = sign
final_conf = conf
if len(prediction_buffer) >= 6:
final_sign = Counter(prediction_buffer).most_common(1)[0][0]
try:
final_conf = probs[label_encoder_classes.index(final_sign)].item()
except:
pass
# Make prediction
input_tensor = torch.FloatTensor(features).unsqueeze(0)
output = model(input_tensor)
probabilities = torch.softmax(output, dim=1)[0]
color = (0, 255, 100) if final_conf > 0.75 else (0, 220, 220)
cv2.putText(image, f"Sign: {final_sign}", (25, 60),
cv2.FONT_HERSHEY_SIMPLEX, 1.8, color, 4)
cv2.putText(image, f"Conf: {final_conf:.1%}", (25, 110),
cv2.FONT_HERSHEY_SIMPLEX, 0.9, (220, 220, 220), 2)
# Get top prediction
predicted_idx = torch.argmax(probabilities).item()
confidence = probabilities[predicted_idx].item()
predicted_sign = label_encoder.inverse_transform([predicted_idx])[0]
top3_p, top3_i = torch.topk(probs, 3)
for i, (p, idx) in enumerate(zip(top3_p, top3_i)):
s = label_encoder_classes[idx.item()]
cv2.putText(image, f"{i + 1}. {s:<18} {p:.1%}",
(25, 155 + i * 40), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (200, 200, 200), 2)
# Add to buffer for smoothing
if confidence > 0.3: # Only add if confident enough
prediction_buffer.append(predicted_sign)
else:
if len(sequence_buffer) < 25:
sequence_buffer.clear()
cv2.putText(image, "No hand detected", (25, 60),
cv2.FONT_HERSHEY_SIMPLEX, 1.3, (0, 0, 255), 3)
# Get smoothed prediction (most common in buffer)
if len(prediction_buffer) >= 5:
from collections import Counter
cv2.putText(image, "ESC:quit SPACE:clear H:landmarks",
(w - 480, h - 25), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (180, 180, 180), 1)
smoothed_sign = Counter(prediction_buffer).most_common(1)[0][0]
else:
smoothed_sign = predicted_sign
cv2.imshow("ASL Recognition", image)
# Get top 5 predictions
top5_prob, top5_idx = torch.topk(probabilities, min(5, num_classes))
key = cv2.waitKey(1) & 0xFF
if key == 27:
break
elif key == 32:
sequence_buffer.clear()
prediction_buffer.clear()
print("Buffers cleared")
elif key in (ord('h'), ord('H')):
show_landmarks = not show_landmarks
print(f"Landmarks display: {'ON' if show_landmarks else 'OFF'}")
# Display prediction area (dark semi-transparent overlay)
overlay = image.copy()
cv2.rectangle(overlay, (10, 10), (500, 280), (0, 0, 0), -1)
cv2.addWeighted(overlay, 0.7, image, 0.3, 0, image)
# Display main prediction
cv2.putText(image, f"Sign: {smoothed_sign}",
(20, 50), cv2.FONT_HERSHEY_SIMPLEX, 1.5, (0, 255, 0), 3)
cv2.putText(image, f"Confidence: {confidence:.1%}",
(20, 90), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 255), 2)
# Display top 5 predictions
cv2.putText(image, "Top 5:",
(20, 130), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2)
y_offset = 160
for i, (prob, idx) in enumerate(zip(top5_prob, top5_idx)):
sign = label_encoder.inverse_transform([idx.item()])[0]
prob_val = prob.item()
# Color code by confidence
if i == 0:
color = (0, 255, 0) # Green for top
elif prob_val > 0.1:
color = (0, 255, 255) # Yellow for decent confidence
else:
color = (128, 128, 128) # Gray for low confidence
cv2.putText(image, f"{i + 1}. {sign}: {prob_val:.1%}",
(30, y_offset), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)
y_offset += 30
else:
# No hand detected
cv2.putText(image, "No hand detected",
(20, 50), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
prediction_buffer.clear()
# Display FPS and info
info_y = image.shape[0] - 60
cv2.putText(image, f"FPS: {current_fps:.1f}",
(20, info_y), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2)
cv2.putText(image, f"Frame: {frame_count}",
(20, info_y + 25), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2)
# Display controls at bottom right
controls_text = "ESC: Exit | SPACE: Clear | H: Landmarks"
text_size = cv2.getTextSize(controls_text, cv2.FONT_HERSHEY_SIMPLEX, 0.5, 1)[0]
cv2.putText(image, controls_text,
(image.shape[1] - text_size[0] - 10, image.shape[0] - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (200, 200, 200), 1)
# Show the image
cv2.imshow('ASL Recognition - Transformer+CNN', image)
# Handle key presses
key = cv2.waitKey(1) & 0xFF
if key == 27: # ESC
print("Exiting...")
break
elif key == 32: # SPACE
prediction_buffer.clear()
print("Prediction buffer cleared")
elif key == ord('h') or key == ord('H'):
show_landmarks = not show_landmarks
print(f"Hand landmarks: {'ON' if show_landmarks else 'OFF'}")
# Cleanup
cap.release()
cv2.destroyAllWindows()
landmarker.close()
print("Recognition stopped.")

View File

@@ -35,188 +35,222 @@ else:
print("=" * 60)
# ===============================
# SELECTED LANDMARK INDICES
# ===============================
IMPORTANT_FACE_INDICES = sorted(list(set([
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
55, 65, 66, 105, 107, 336, 296, 334,
33, 133, 160, 159, 158, 144, 145, 153,
362, 263, 387, 386, 385, 373, 374, 380,
1, 2, 98, 327,
61, 185, 40, 39, 37, 0, 267, 269, 270, 409,
291, 146, 91, 181, 84, 17, 314, 405, 321, 375,
78, 191, 80, 81, 82, 13, 312, 311, 310, 415,
308, 324, 318, 402, 317, 14, 87, 178, 88, 95
])))
NUM_FACE_POINTS = len(IMPORTANT_FACE_INDICES)
NUM_HAND_POINTS = 21 * 2
TOTAL_POINTS_PER_FRAME = NUM_HAND_POINTS + NUM_FACE_POINTS
# ===============================
# DATA LOADING - HANDLES PARTIAL NaN
# ENHANCED DATA EXTRACTION
# ===============================
def load_kaggle_asl_data(base_path):
train_df = pd.read_csv(os.path.join(base_path, "train.csv"))
with open(os.path.join(base_path, "sign_to_prediction_index_map.json")) as f:
sign_to_idx = json.load(f)
return train_df, sign_to_idx
def extract_hand_landmarks_from_parquet(path):
"""Extract hand landmarks - ONLY uses frames with valid (non-NaN) data"""
def extract_multi_landmarks(path, min_valid_frames=5):
"""
Extract both hands + selected face landmarks with modality flags
Returns: dict with 'landmarks', 'left_hand_valid', 'right_hand_valid', 'face_valid'
"""
try:
df = pd.read_parquet(path)
seq = []
left_valid_frames = []
right_valid_frames = []
face_valid_frames = []
# Get hand data
left = df[df["type"] == "left_hand"]
right = df[df["type"] == "right_hand"]
if len(left) == 0 and len(right) == 0:
all_types = df["type"].unique()
if "left_hand" in all_types or "right_hand" in all_types or "face" in all_types:
frames = sorted(df["frame"].unique())
else:
return None
# Count valid (non-NaN) rows for each hand
left_valid = 0
right_valid = 0
if len(left) > 0:
left_valid = left[['x', 'y', 'z']].notna().all(axis=1).sum()
if len(right) > 0:
right_valid = right[['x', 'y', 'z']].notna().all(axis=1).sum()
# No valid data at all
if left_valid == 0 and right_valid == 0:
if frames is None or len(frames) < min_valid_frames:
return None
# Choose hand with more valid data
hand = left if left_valid >= right_valid else right
# Get unique frames
frames = sorted(hand['frame'].unique())
landmarks_seq = []
for frame in frames:
lm_frame = hand[hand['frame'] == frame]
frame_df = df[df["frame"] == frame]
frame_points = np.full((TOTAL_POINTS_PER_FRAME, 3), np.nan, dtype=np.float32)
# Count how many valid landmarks this frame has
valid_count = lm_frame[['x', 'y', 'z']].notna().all(axis=1).sum()
pos = 0
left_valid = False
right_valid = False
face_valid = False
# Skip frames with too few valid landmarks
if valid_count < 10:
continue
# Left hand
left = frame_df[frame_df["type"] == "left_hand"]
if len(left) >= 15:
valid_count = 0
for i in range(21):
row = left[left["landmark_index"] == i]
if len(row) > 0 and row[['x', 'y', 'z']].notna().all().all():
frame_points[pos] = row[['x', 'y', 'z']].values[0]
valid_count += 1
pos += 1
left_valid = (valid_count >= 15)
else:
pos += 21
# Extract landmarks for this frame
frame_landmarks = []
valid_landmarks_in_frame = 0
# Right hand
right = frame_df[frame_df["type"] == "right_hand"]
if len(right) >= 15:
valid_count = 0
for i in range(21):
row = right[right["landmark_index"] == i]
if len(row) > 0 and row[['x', 'y', 'z']].notna().all().all():
frame_points[pos] = row[['x', 'y', 'z']].values[0]
valid_count += 1
pos += 1
right_valid = (valid_count >= 15)
else:
pos += 21
for i in range(21):
lm = lm_frame[lm_frame['landmark_index'] == i]
# Face
face = frame_df[frame_df["type"] == "face"]
if len(face) > 0:
valid_count = 0
for idx in IMPORTANT_FACE_INDICES:
row = face[face["landmark_index"] == idx]
if len(row) > 0 and row[['x', 'y', 'z']].notna().all().all():
frame_points[pos] = row[['x', 'y', 'z']].values[0]
valid_count += 1
pos += 1
face_valid = (valid_count >= len(IMPORTANT_FACE_INDICES) * 0.5)
if len(lm) == 0:
frame_landmarks.append([0.0, 0.0, 0.0])
else:
x = float(lm['x'].iloc[0])
y = float(lm['y'].iloc[0])
z = float(lm['z'].iloc[0])
valid_ratio = 1 - np.isnan(frame_points).mean()
if valid_ratio >= 0.40:
frame_points = np.nan_to_num(frame_points, nan=0.0)
seq.append(frame_points)
left_valid_frames.append(left_valid)
right_valid_frames.append(right_valid)
face_valid_frames.append(face_valid)
# Check if valid
if pd.notna(x) and pd.notna(y) and pd.notna(z):
frame_landmarks.append([x, y, z])
valid_landmarks_in_frame += 1
else:
frame_landmarks.append([0.0, 0.0, 0.0])
# Only add frame if it has enough valid landmarks
if valid_landmarks_in_frame >= 10:
landmarks_seq.append(frame_landmarks)
# Need at least 3 valid frames
if len(landmarks_seq) < 3:
if len(seq) < min_valid_frames:
return None
return np.array(landmarks_seq, dtype=np.float32)
return {
'landmarks': np.stack(seq),
'left_hand_valid': np.array(left_valid_frames),
'right_hand_valid': np.array(right_valid_frames),
'face_valid': np.array(face_valid_frames)
}
except Exception as e:
except Exception:
return None
def get_features_sequence(landmarks_seq, max_frames=100):
"""Extract features from landmark sequence"""
if landmarks_seq is None or len(landmarks_seq) == 0:
return None, None
# Center on wrist (landmark 0)
wrist = landmarks_seq[:, 0:1, :].copy()
landmarks_seq = landmarks_seq - wrist
# Scale normalization using wrist to middle finger MCP (landmark 9)
scale = np.linalg.norm(landmarks_seq[:, 9, :] - np.zeros(3), axis=1, keepdims=True)
scale = np.maximum(scale, 1e-6) # Avoid division by zero
landmarks_seq = landmarks_seq / scale[:, np.newaxis, :]
# Clean up any remaining NaN/Inf
landmarks_seq = np.nan_to_num(landmarks_seq, nan=0.0, posinf=0.0, neginf=0.0)
# Clip extreme values
landmarks_seq = np.clip(landmarks_seq, -10, 10)
# Calculate finger curl features
tips = [4, 8, 12, 16, 20] # Thumb, index, middle, ring, pinky tips
bases = [1, 5, 9, 13, 17] # Corresponding base joints
curl_features = []
for b, t in zip(bases, tips):
curl = np.linalg.norm(landmarks_seq[:, t] - landmarks_seq[:, b], axis=1)
curl_features.append(curl)
curl_features = np.stack(curl_features, axis=1) # (T, 5)
# Temporal deltas (motion)
deltas = np.zeros_like(landmarks_seq)
if len(landmarks_seq) > 1:
deltas[1:] = landmarks_seq[1:] - landmarks_seq[:-1]
# Flatten each component separately, then concatenate
landmarks_flat = landmarks_seq.reshape(landmarks_seq.shape[0], -1) # (T, 63)
deltas_flat = deltas.reshape(deltas.shape[0], -1) # (T, 63)
# curl_features is already (T, 5)
# Combine: 63 + 63 + 5 = 131 features per frame
seq = np.concatenate([
landmarks_flat,
deltas_flat,
curl_features
], axis=1)
# Pad or truncate to max_frames
T, F = seq.shape
if T < max_frames:
# Pad with zeros
pad = np.zeros((max_frames - T, F), dtype=np.float32)
seq = np.concatenate([seq, pad], axis=0)
elif T > max_frames:
# Truncate
seq = seq[:max_frames, :]
# Create attention mask (True for valid positions)
valid_mask = np.zeros(max_frames, dtype=bool)
valid_mask[:min(T, max_frames)] = True
return seq.astype(np.float32), valid_mask
def process_row(row, base_path, max_frames=100):
"""Process a single row - worker function for multiprocessing"""
path = os.path.join(base_path, row["path"])
if not os.path.exists(path):
def get_features_sequence(landmarks_data, max_frames=100):
"""
Enhanced feature extraction with separate modality processing
"""
if landmarks_data is None:
return None, None, None
landmarks_3d = landmarks_data['landmarks']
if len(landmarks_3d) == 0:
return None, None, None
T, N, _ = landmarks_3d.shape
# Separate modalities for independent normalization
left_hand = landmarks_3d[:, :21, :]
right_hand = landmarks_3d[:, 21:42, :]
face = landmarks_3d[:, 42:, :]
# Independent centering per modality
features_list = []
for modality, valid_mask in [
(left_hand, landmarks_data['left_hand_valid']),
(right_hand, landmarks_data['right_hand_valid']),
(face, landmarks_data['face_valid'])
]:
# Center on modality-specific mean
valid_frames = modality[valid_mask] if valid_mask.any() else modality
if len(valid_frames) > 0:
center = np.mean(valid_frames, axis=(0, 1), keepdims=True)
spread = np.std(valid_frames, axis=(0, 1), keepdims=True).max()
else:
center = 0
spread = 1
modality_norm = (modality - center) / max(spread, 1e-6)
flat = modality_norm.reshape(T, -1)
# Deltas
deltas = np.zeros_like(flat)
if T > 1:
deltas[1:] = flat[1:] - flat[:-1]
features_list.append(flat)
features_list.append(deltas)
# Combine all modalities
features = np.concatenate(features_list, axis=1)
# Create modality availability mask (which frames have which modalities)
modality_mask = np.stack([
landmarks_data['left_hand_valid'],
landmarks_data['right_hand_valid'],
landmarks_data['face_valid']
], axis=1).astype(np.float32) # (T, 3)
# Pad/truncate
if T < max_frames:
pad = np.zeros((max_frames - T, features.shape[1]), dtype=np.float32)
features = np.concatenate([features, pad], axis=0)
mask_pad = np.zeros((max_frames - T, 3), dtype=np.float32)
modality_mask = np.concatenate([modality_mask, mask_pad], axis=0)
frame_mask = np.zeros(max_frames, dtype=bool)
frame_mask[:T] = True
else:
features = features[:max_frames]
modality_mask = modality_mask[:max_frames]
frame_mask = np.ones(max_frames, dtype=bool)
features = np.nan_to_num(features, nan=0.0, posinf=0.0, neginf=0.0)
features = np.clip(features, -30, 30)
return features.astype(np.float32), frame_mask, modality_mask
def process_row(row_data, base_path, max_frames=100):
"""Process a single row - expects tuple of (path, sign)"""
path_rel, sign = row_data
path = os.path.join(base_path, path_rel)
if not os.path.exists(path):
return None, None, None, None
try:
# Extract landmarks
lm = extract_hand_landmarks_from_parquet(path)
if lm is None:
return None, None, None
lm_data = extract_multi_landmarks(path)
if lm_data is None:
return None, None, None, None
# Get features
feat, mask = get_features_sequence(lm, max_frames)
feat, frame_mask, modality_mask = get_features_sequence(lm_data, max_frames)
if feat is None:
return None, None, None
return None, None, None, None
# Final safety check
if np.isnan(feat).any() or np.isinf(feat).any():
return None, None, None
return feat, frame_mask, modality_mask, sign
return feat, mask, row["sign"]
except Exception as e:
return None, None, None
except Exception:
return None, None, None, None
# ===============================
# TRANSFORMER MODEL
# ENHANCED MODEL WITH MODALITY AWARENESS
# ===============================
class PositionalEncoding(nn.Module):
def __init__(self, d_model, max_len=128):
@@ -232,16 +266,19 @@ class PositionalEncoding(nn.Module):
return x + self.pe[:, :x.size(1)]
class TransformerASL(nn.Module):
def __init__(self, input_dim, num_classes, d_model=256, nhead=8, num_layers=4):
class ModalityAwareTransformer(nn.Module):
def __init__(self, input_dim, num_classes, d_model=384, nhead=8, num_layers=5):
super().__init__()
# Input projection
# Main projection
self.proj = nn.Linear(input_dim, d_model)
self.norm_in = nn.LayerNorm(d_model)
self.pos = PositionalEncoding(d_model, max_len=128)
# Transformer encoder
# Modality embedding (3 modalities: left_hand, right_hand, face)
self.modality_embed = nn.Linear(3, d_model)
self.norm_in = nn.LayerNorm(d_model)
self.pos = PositionalEncoding(d_model)
enc_layer = nn.TransformerEncoderLayer(
d_model=d_model,
nhead=nhead,
@@ -253,30 +290,45 @@ class TransformerASL(nn.Module):
)
self.encoder = nn.TransformerEncoder(enc_layer, num_layers=num_layers)
# Classification head
self.head = nn.Sequential(
nn.LayerNorm(d_model),
nn.Dropout(0.25),
nn.Linear(d_model, num_classes)
)
def forward(self, x, key_padding_mask=None):
# x: (batch, seq_len, input_dim)
# key_padding_mask: (batch, seq_len) - True for padding positions
def forward(self, x, modality_mask=None, key_padding_mask=None):
# Project features
x = self.proj(x)
# Add modality information if available
if modality_mask is not None:
mod_embed = self.modality_embed(modality_mask)
x = x + mod_embed
x = self.norm_in(x)
x = self.pos(x)
x = self.encoder(x, src_key_padding_mask=key_padding_mask)
# Global average pooling over valid positions
x = x.mean(dim=1)
# Weighted average (giving more weight to frames with more valid modalities)
if modality_mask is not None:
weights = modality_mask.sum(dim=-1, keepdim=True) + 1e-6 # (B, T, 1)
weights = weights / weights.sum(dim=1, keepdim=True)
x = (x * weights).sum(dim=1)
else:
x = x.mean(dim=1)
return self.head(x)
def load_kaggle_asl_data(base_path):
"""Load training metadata"""
train_path = os.path.join(base_path, "train.csv")
train_df = pd.read_csv(train_path)
return train_df, None
# ===============================
# MAIN TRAINING
# MAIN
# ===============================
def main():
base_path = "asl_kaggle"
@@ -284,163 +336,142 @@ def main():
MIN_SAMPLES_PER_CLASS = 5
print("\nLoading metadata...")
train_df, sign_to_idx = load_kaggle_asl_data(base_path)
print(f"Total sequences: {len(train_df)}")
train_df, _ = load_kaggle_asl_data(base_path)
rows = [row for _, row in train_df.iterrows()]
# Convert to simple tuples for multiprocessing compatibility
rows = [(row['path'], row['sign']) for _, row in train_df.iterrows()]
print("\nProcessing sequences (this will take a few minutes)...")
print("Expected: ~36,000 valid sequences based on diagnostic")
# Process with multiprocessing
print("\nProcessing sequences with BOTH hands + FACE (enhanced)...")
with Pool(cpu_count()) as pool:
results = list(tqdm(
pool.imap(
partial(process_row, base_path=base_path, max_frames=max_frames),
rows,
chunksize=100
chunksize=80
),
total=len(rows),
desc="Extracting landmarks"
desc="Landmarks extraction"
))
# Filter valid results
X_list, masks_list, y_list = [], [], []
for feat, mask, sign in results:
if feat is not None and mask is not None and sign is not None:
if feat.shape[0] == max_frames:
X_list.append(feat)
masks_list.append(mask)
y_list.append(sign)
X_list, frame_masks_list, modality_masks_list, y_list = [], [], [], []
for feat, frame_mask, modality_mask, sign in results:
if feat is not None and frame_mask is not None:
X_list.append(feat)
frame_masks_list.append(frame_mask)
modality_masks_list.append(modality_mask)
y_list.append(sign)
print(f"\n✓ Successfully extracted: {len(X_list)} valid sequences")
print(f" Success rate: {len(X_list) / len(train_df) * 100:.1f}%")
if len(X_list) < 100:
print("❌ Too few valid sequences found!")
print(" This shouldn't happen - please share this output for debugging")
if not X_list:
print("No valid sequences extracted!")
return
# Stack into arrays
X = np.stack(X_list)
masks = np.stack(masks_list)
frame_masks = np.stack(frame_masks_list)
modality_masks = np.stack(modality_masks_list)
print(f"\nData shape: {X.shape}")
print(f"Feature dimension: {X.shape[2]}")
print(f"\nExtracted {len(X):,} sequences")
print(f"Feature shape: {X.shape[1:]} (input_dim = {X.shape[2]})")
print(f"Modality mask shape: {modality_masks.shape}")
# Global normalization
print("Normalizing features...")
X = np.clip(X, -10.0, 10.0)
X = np.clip(X, -30, 30)
mean = X.mean(axis=(0, 1), keepdims=True)
std = X.std(axis=(0, 1), keepdims=True) + 1e-8
X = (X - mean) / std
# Encode labels
# Labels
le = LabelEncoder()
y = le.fit_transform(y_list)
# Filter classes with too few samples
# Filter rare classes
counts = Counter(y)
valid_classes = [cls for cls, cnt in counts.items() if cnt >= MIN_SAMPLES_PER_CLASS]
mask_valid = np.isin(y, valid_classes)
valid = [k for k, v in counts.items() if v >= MIN_SAMPLES_PER_CLASS]
mask = np.isin(y, valid)
X = X[mask_valid]
masks = masks[mask_valid]
y = y[mask_valid]
X = X[mask]
frame_masks = frame_masks[mask]
modality_masks = modality_masks[mask]
y = y[mask]
# Re-encode after filtering
le = LabelEncoder()
y = le.fit_transform(y)
print(f"\nFinal dataset after filtering:")
print(f" Samples: {len(X):,}")
print(f" Classes: {len(le.classes_)}")
print(f" Sign examples: {list(le.classes_[:10])}")
print(f"After filtering: {len(X):,} samples | {len(le.classes_)} classes")
# Train-test split
X_train, X_test, masks_train, masks_test, y_train, y_test = train_test_split(
X, masks, y, test_size=0.15, stratify=y, random_state=42
# Analyze modality usage
print("\nModality statistics:")
print(f" Sequences with left hand: {(modality_masks[:, :, 0].sum(axis=1) > 0).mean() * 100:.1f}%")
print(f" Sequences with right hand: {(modality_masks[:, :, 1].sum(axis=1) > 0).mean() * 100:.1f}%")
print(f" Sequences with face: {(modality_masks[:, :, 2].sum(axis=1) > 0).mean() * 100:.1f}%")
# Split
X_tr, X_te, fm_tr, fm_te, mm_tr, mm_te, y_tr, y_te = train_test_split(
X, frame_masks, modality_masks, y, test_size=0.15, stratify=y, random_state=42
)
print(f"\nTrain set: {len(X_train):,} samples")
print(f"Test set: {len(X_test):,} samples")
# Dataset wrapper
class ASLSequenceDataset(Dataset):
def __init__(self, X, masks, y):
# Dataset
class ASLMultiDataset(Dataset):
def __init__(self, X, frame_masks, modality_masks, y):
self.X = torch.from_numpy(X).float()
self.masks = torch.from_numpy(masks)
self.frame_masks = torch.from_numpy(frame_masks).bool()
self.modality_masks = torch.from_numpy(modality_masks).float()
self.y = torch.from_numpy(y).long()
def __len__(self):
return len(self.X)
def __getitem__(self, idx):
return self.X[idx], self.masks[idx], self.y[idx]
return self.X[idx], self.frame_masks[idx], self.modality_masks[idx], self.y[idx]
# DataLoaders
batch_size = 128 if device.type == 'cuda' else 64
batch_size = 64 if device.type == 'cuda' else 32
train_loader = DataLoader(
ASLSequenceDataset(X_train, masks_train, y_train),
batch_size=batch_size,
shuffle=True,
num_workers=4,
pin_memory=True if device.type == 'cuda' else False
ASLMultiDataset(X_tr, fm_tr, mm_tr, y_tr),
batch_size=batch_size, shuffle=True,
num_workers=4, pin_memory=device.type == 'cuda'
)
test_loader = DataLoader(
ASLSequenceDataset(X_test, masks_test, y_test),
batch_size=batch_size * 2,
shuffle=False,
num_workers=4,
pin_memory=True if device.type == 'cuda' else False
ASLMultiDataset(X_te, fm_te, mm_te, y_te),
batch_size=batch_size * 2, shuffle=False,
num_workers=4, pin_memory=device.type == 'cuda'
)
# Initialize model
print("\nInitializing model...")
model = TransformerASL(
# Enhanced model
model = ModalityAwareTransformer(
input_dim=X.shape[2],
num_classes=len(le.classes_),
d_model=256,
d_model=384,
nhead=8,
num_layers=4
num_layers=5
).to(device)
total_params = sum(p.numel() for p in model.parameters())
print(f"Model parameters: {total_params:,}")
print(f"\nModel parameters: {sum(p.numel() for p in model.parameters()):,}")
# Training setup
criterion = nn.CrossEntropyLoss(label_smoothing=0.05)
optimizer = optim.AdamW(model.parameters(), lr=5e-4, weight_decay=1e-4)
scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10, T_mult=2)
optimizer = optim.AdamW(model.parameters(), lr=4e-4, weight_decay=1e-4)
scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10)
# Training loop
best_acc = 0.0
epochs = 60
print("\n" + "=" * 60)
print("STARTING TRAINING")
print("=" * 60)
epochs = 70
for epoch in range(epochs):
# Train
model.train()
total_loss = 0
correct = total = 0
total_loss = correct = total = 0
for x, mask, yb in tqdm(train_loader, desc=f"Epoch {epoch + 1}/{epochs}", leave=False):
x, mask, yb = x.to(device), mask.to(device), yb.to(device)
for x, frame_mask, modality_mask, yb in tqdm(train_loader, desc=f"Epoch {epoch + 1}", leave=False):
x = x.to(device)
frame_mask = frame_mask.to(device)
modality_mask = modality_mask.to(device)
yb = yb.to(device)
# Invert mask: True for padding positions
key_mask = ~mask
key_padding_mask = ~frame_mask
optimizer.zero_grad(set_to_none=True)
logits = model(x, key_padding_mask=key_mask)
logits = model(x, modality_mask=modality_mask, key_padding_mask=key_padding_mask)
loss = criterion(logits, yb)
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
optimizer.step()
total_loss += loss.item()
@@ -449,49 +480,35 @@ def main():
train_acc = correct / total * 100
# Evaluate
# Eval
model.eval()
correct = total = 0
with torch.no_grad():
for x, mask, yb in test_loader:
x, mask, yb = x.to(device), mask.to(device), yb.to(device)
key_mask = ~mask
logits = model(x, key_padding_mask=key_mask)
for x, frame_mask, modality_mask, yb in test_loader:
x = x.to(device)
frame_mask = frame_mask.to(device)
modality_mask = modality_mask.to(device)
yb = yb.to(device)
key_padding_mask = ~frame_mask
logits = model(x, modality_mask=modality_mask, key_padding_mask=key_padding_mask)
correct += (logits.argmax(-1) == yb).sum().item()
total += yb.size(0)
test_acc = correct / total * 100
scheduler.step()
# Print progress
print(f"[{epoch + 1:2d}/{epochs}] Loss: {total_loss / len(train_loader):.4f} | "
f"Train: {train_acc:.2f}% | Test: {test_acc:.2f}%", end="")
scheduler.step()
# Save best model
if test_acc > best_acc:
best_acc = test_acc
torch.save({
'model': model.state_dict(),
'optimizer': optimizer.state_dict(),
'label_encoder_classes': le.classes_,
'acc': best_acc,
'epoch': epoch,
'input_dim': X.shape[2],
'num_classes': len(le.classes_),
'd_model': 256,
'nhead': 8,
'num_layers': 4
}, "best_asl_transformer.pth")
print(f" → New best: {best_acc:.2f}% ✓")
torch.save(model.state_dict(), "best_asl_modality_aware.pth")
print(" → saved")
else:
print()
print("\n" + "=" * 60)
print(f"✓ Training complete!")
print(f"✓ Best test accuracy: {best_acc:.2f}%")
print(f"✓ Model saved: best_asl_transformer.pth")
print("=" * 60)
print(f"\nBest test accuracy: {best_acc:.2f}%")
if __name__ == "__main__":