File size: 38,354 Bytes
e2cffd9 76ef804 e2cffd9 76ef804 e2cffd9 76ef804 e2cffd9 76ef804 eea2701 76ef804 4c95ba8 76ef804 4c95ba8 76ef804 4c95ba8 76ef804 4c95ba8 76ef804 4c95ba8 ec0f44e 76ef804 e2cffd9 76ef804 e2cffd9 76ef804 bbe5030 76ef804 bbe5030 4c95ba8 bbe5030 76ef804 e2cffd9 76ef804 e2cffd9 76ef804 e2cffd9 eea2701 e2cffd9 76ef804 e2cffd9 76ef804 e2cffd9 76ef804 e2cffd9 6739284 e2cffd9 ec0f44e e2cffd9 ec0f44e e2cffd9 bbe5030 e2cffd9 bbe5030 4c95ba8 bbe5030 e2cffd9 ec0f44e e2cffd9 ec0f44e e2cffd9 bbe5030 e2cffd9 4c56f90 e2cffd9 4c56f90 e2cffd9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 |
"""
ISL Sign Language Translation - TechMatrix Solvers Initiative
Main Streamlit Application
Developed by: TechMatrix Solvers Team
- Abhay Gupta (Team Lead)
- Kripanshu Gupta (Backend Developer)
- Dipanshu Patel (UI/UX Designer)
- Bhumika Patel (Deployment & Female Presenter)
Institution: Shri Ram Group of Institutions
"""
import streamlit as st
# Configure Streamlit page first
st.set_page_config(
page_title="ISL Translation - TechMatrix Solvers",
page_icon="🤟",
layout="wide",
initial_sidebar_state="expanded"
)
# Show loading message
st.write("🚀 TechMatrix Solvers ISL Translator Loading...")
# Import dependencies with error handling
try:
import os
# Set environment variables for better compatibility
os.environ["KERAS_BACKEND"] = "tensorflow"
os.environ["HF_HOME"] = "/tmp/huggingface"
os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers"
# Core imports
import numpy as np
import pandas as pd
import tempfile
import time
from PIL import Image
import subprocess
from typing import NamedTuple
import json
import shutil
import platform
import uuid
# Try OpenCV import
try:
import cv2
except Exception as cv_error:
st.warning(f"OpenCV import issue: {cv_error}")
cv2 = None
# Try ML library imports
try:
import keras
from keras.models import Sequential
from keras.layers import LSTM, Dense, Bidirectional, Dropout, Input, BatchNormalization
except Exception as keras_error:
st.warning(f"Keras import issue: {keras_error}")
keras = None
# Try video processing
try:
import ffmpeg
except Exception as ffmpeg_error:
st.warning(f"FFmpeg import issue: {ffmpeg_error}")
ffmpeg = None
# Try HuggingFace Hub
try:
from huggingface_hub import hf_hub_download
except Exception as hf_error:
st.warning(f"HuggingFace Hub import issue: {hf_error}")
hf_hub_download = None
# Try custom modules with fallback - import each module separately for better error handling
pose_models = None
expression_mapping = None
isl_processor = None
utils = None
# Try importing pose_models
try:
from pose_models import create_bodypose_model, create_handpose_model
pose_models = True
st.success("✅ Pose models imported successfully")
except Exception as pose_error:
st.warning(f"Pose models import issue: {pose_error}")
pose_models = None
# Try importing expression_mapping
try:
from expression_mapping import expression_mapping
st.success("✅ Expression mapping imported successfully")
except Exception as expr_error:
st.warning(f"Expression mapping import issue: {expr_error}")
# Fallback expression mapping
expression_mapping = {
'hello': 0, 'thank_you': 1, 'please': 2, 'sorry': 3, 'help': 4,
'good': 5, 'bad': 6, 'yes': 7, 'no': 8, 'water': 9,
'food': 10, 'home': 11, 'work': 12, 'school': 13, 'family': 14
}
# Try importing ISL processor
try:
from isl_processor import ISLTranslationModel
isl_processor = True
st.success("✅ ISL processor imported successfully")
except Exception as isl_error:
st.warning(f"ISL processor import issue: {isl_error}")
isl_processor = None
# Try importing pose_utils
try:
import pose_utils as utils
st.success("✅ Pose utils imported successfully")
except Exception as utils_error:
st.warning(f"Pose utils import issue: {utils_error}")
utils = None
st.success("✅ Core dependencies loaded successfully!")
except ImportError as e:
st.error(f"❌ Critical import error: {e}")
st.error("Running in fallback mode with limited functionality.")
# Ensure we have utils available globally after the main import block
if utils is None:
try:
import pose_utils as utils
st.info("ℹ️ Pose utils loaded on secondary attempt")
except ImportError as utils_error:
st.error(f"❌ Failed to import pose_utils: {utils_error}")
utils = None
# Ensure expression_mapping is available and create index-to-label mapping
if expression_mapping is None:
st.warning("⚠️ Using fallback expression mapping")
expression_mapping = {
0: 'hello', 1: 'thank_you', 2: 'please', 3: 'sorry', 4: 'help',
5: 'good', 6: 'bad', 7: 'yes', 8: 'no', 9: 'water',
10: 'food', 11: 'home', 12: 'work', 13: 'school', 14: 'family'
}
# Create index-to-label mapping function for safe access
def get_sign_label(index):
"""Safely get sign label from prediction index"""
if isinstance(expression_mapping, dict):
return expression_mapping.get(int(index), f'unknown_sign_{index}')
else:
return f'sign_{index}'
# System information will be shown in About section
class VideoProbeResult(NamedTuple):
"""Structure for video probe results"""
return_code: int
json: str
error: str
def probe_video_info(file_path) -> VideoProbeResult:
"""
Probe video file for metadata using FFprobe
Args:
file_path: Path to video file
Returns:
VideoProbeResult containing metadata
"""
command_array = [
"ffprobe",
"-v", "quiet",
"-print_format", "json",
"-show_format",
"-show_streams",
file_path
]
result = subprocess.run(
command_array,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
universal_newlines=True
)
return VideoProbeResult(
return_code=result.returncode,
json=result.stdout,
error=result.stderr
)
# Define feature columns for time series processing
body_features = [f'bodypeaks_x_{i}' for i in range(15)] + [f'bodypeaks_y_{i}' for i in range(15)]
hand0_features = [f'hand0peaks_x_{i}' for i in range(21)] + [f'hand0peaks_y_{i}' for i in range(21)] + [f'hand0peaks_peaktxt{i}' for i in range(21)]
hand1_features = [f'hand1peaks_x_{i}' for i in range(21)] + [f'hand1peaks_y_{i}' for i in range(21)] + [f'hand1peaks_peaktxt{i}' for i in range(21)]
feature_columns_processed = body_features + hand0_features + hand1_features
label_columns = ['Expression_encoded']
@st.cache_resource
def create_time_series_sequences(isl_data, feature_columns, label_columns, window_size=20):
"""
Creates time series sequences from DataFrame with specified window size
Args:
isl_data: Input DataFrame with ISL data
feature_columns: List of feature column names
label_columns: List of label column names
window_size: Size of temporal window for sequence creation
Returns:
tuple: (X_sequences, y_sequences) for training/inference
"""
if isl_data.empty:
return [], []
X_sequences = []
y_sequences = []
for group, file_df in isl_data.groupby(['Type', 'Expression_encoded', 'FileName']):
expr_type, expression, filename = group
# Create blank frame for padding
blank_frame = np.zeros((1, 156))
for idx, window_data in enumerate([file_df[i:i+window_size] for i in range(0, file_df.shape[0], 1)]):
if window_data.shape[0] < window_size:
# Pad sequence with blank frames at the beginning
padding_needed = window_size - window_data.shape[0]
padded_sequence = np.concatenate(
(np.repeat(blank_frame, padding_needed, axis=0),
window_data[feature_columns].values),
axis=0
)
X_sequences.append(padded_sequence)
y_sequences.append(expression)
continue
X_sequences.append(window_data[feature_columns].values)
y_sequences.append(expression)
return X_sequences, y_sequences
# Global translation model variable
translation_model = None
@st.cache_resource
def load_translation_model():
"""
Load and configure the LSTM translation model
Returns:
Configured Keras Sequential model for ISL translation or None if failed
"""
try:
if keras is None or hf_hub_download is None:
st.warning("Keras or HuggingFace Hub not available. Model loading skipped.")
return None
# Download pre-trained model file
model_file = hf_hub_download(
repo_id="sunilsarolkar/isl-translation-model",
filename="isl_model_final.keras"
)
# Try to load the complete model first
try:
model = keras.models.load_model(model_file)
st.success("✅ Model loaded successfully from saved file")
return model
except Exception as load_error:
st.warning(f"Failed to load complete model: {load_error}")
st.info("Attempting to build model architecture and load weights...")
# Fallback: Build model architecture and load weights
model = Sequential()
model.add(Input(shape=((20, 156))))
model.add(keras.layers.Masking(mask_value=0.))
model.add(BatchNormalization())
model.add(Bidirectional(LSTM(32, recurrent_dropout=0.2, return_sequences=True)))
model.add(Dropout(0.2))
model.add(Bidirectional(LSTM(32, recurrent_dropout=0.2)))
model.add(keras.layers.Activation('elu'))
model.add(Dense(32, use_bias=False, kernel_initializer='he_normal'))
model.add(BatchNormalization())
model.add(Dropout(0.2))
model.add(keras.layers.Activation('elu'))
model.add(Dense(32, kernel_initializer='he_normal', use_bias=False))
model.add(BatchNormalization())
model.add(keras.layers.Activation('elu'))
model.add(Dropout(0.2))
# Determine number of classes - use 167 for the full dataset or fallback size
num_classes = len(list(expression_mapping.keys())) if expression_mapping else 167
model.add(Dense(num_classes, activation='softmax'))
# Try to load weights
model.load_weights(model_file)
st.success("✅ Model architecture built and weights loaded successfully")
return model
except Exception as e:
st.error(f"Failed to load translation model: {e}")
return None
# Load test data
@st.cache_data
def load_test_data():
"""Load test dataset and file information"""
testing_cleaned_path = hf_hub_download(
repo_id="sunilsarolkar/isl-test-data",
filename="testing_cleaned.csv",
repo_type="dataset"
)
test_files_path = hf_hub_download(
repo_id="sunilsarolkar/isl-test-data",
filename="test_files.csv",
repo_type="dataset"
)
testing_df = pd.read_csv(testing_cleaned_path)
test_files_df = pd.read_csv(test_files_path)
return testing_df, test_files_df
# Test data will be loaded when needed
class VideoWriter:
"""Custom video writer using FFmpeg for better compatibility"""
def __init__(self, output_file, input_fps, input_framesize, input_pix_fmt, input_vcodec):
self.ff_process = (
ffmpeg
.input('pipe:',
format='rawvideo',
pix_fmt="bgr24",
s=f'{input_framesize[1]}x{input_framesize[0]}',
r=input_fps)
.output(output_file, pix_fmt=input_pix_fmt, vcodec=input_vcodec)
.overwrite_output()
.run_async(pipe_stdin=True)
)
def write_frame(self, frame):
"""Write a single frame to the video"""
self.ff_process.stdin.write(frame.tobytes())
def close(self):
"""Close the video writer"""
self.ff_process.stdin.close()
self.ff_process.wait()
def calculate_weighted_average(numbers, weights):
"""
Calculate weighted average of numbers
Args:
numbers: List of numbers
weights: List of weights
Returns:
float: Weighted average
"""
if sum(weights) == 0:
return 0
return sum(x * y for x, y in zip(numbers, weights)) / sum(weights)
@st.cache_data
def resize_image(image, width=None, height=None, interpolation=cv2.INTER_AREA):
"""
Resize image maintaining aspect ratio
Args:
image: Input image
width: Target width
height: Target height
interpolation: OpenCV interpolation method
Returns:
Resized image
"""
dimensions = None
(h, w) = image.shape[:2]
if width is None and height is None:
return image
if width is None:
ratio = height / float(h)
dimensions = (int(w * ratio), height)
else:
ratio = width / float(w)
dimensions = (width, int(h * ratio))
resized = cv2.resize(image, dimensions, interpolation=interpolation)
return resized
# Page configuration already set at the top
st.title('🤟 ISL Sign Language Translation - TechMatrix Solvers Initiative')
# Add custom CSS for sidebar styling
st.markdown(
"""
<style>
[data-testid="stSidebar"][aria-expanded="true"] > div:first-child {
width: 350px;
}
[data-testid="stSidebar"][aria-expanded="false"] > div:first-child {
width: 350px;
margin-left: -350px;
}
.team-info {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
padding: 1rem;
border-radius: 0.5rem;
margin: 1rem 0;
}
.tech-matrix-header {
background: linear-gradient(90deg, #1e3a8a, #7c3aed);
color: white;
padding: 1rem;
border-radius: 0.5rem;
text-align: center;
margin-bottom: 1rem;
}
</style>
""",
unsafe_allow_html=True,
)
# Add team branding header
st.markdown(
"""
<div class="tech-matrix-header">
<h2>🚀 TechMatrix Solvers</h2>
<p>Innovating Accessible Technology Solutions</p>
</div>
""",
unsafe_allow_html=True
)
# Sidebar configuration
st.sidebar.title('🤟 ISL Translation System')
st.sidebar.subheader('Configuration')
# Team information in sidebar
st.sidebar.markdown(
"""
<div class="team-info">
<h3>👨💻 Development Team</h3>
<ul>
<li><strong>Abhay Gupta</strong> - Team Lead</li>
<li><strong>Kripanshu Gupta</strong> - Backend Dev</li>
<li><strong>Dipanshu Patel</strong> - UI/UX Designer</li>
<li><strong>Bhumika Patel</strong> - Deployment</li>
</ul>
<p><em>Shri Ram Group of Institutions</em></p>
</div>
""",
unsafe_allow_html=True
)
# Initialize frame-wise outputs storage
frame_predictions = {}
# Application mode selection
app_mode = st.sidebar.selectbox(
'Choose Application Mode',
['About Project', 'Test Video Translation']
)
if app_mode == 'About Project':
st.markdown(
"""
## 🎯 Project Overview
Welcome to the **ISL Sign Language Translation System** developed by **TechMatrix Solvers**.
This cutting-edge application demonstrates real-time Indian Sign Language recognition and
translation using advanced deep learning techniques.
### 🏗️ Technical Architecture
Our system combines multiple state-of-the-art technologies:
1. **Body Pose Estimation**: 25-point skeletal tracking using OpenPose
2. **Hand Landmark Detection**: 21-point hand keypoint identification
3. **Temporal Modeling**: Bidirectional LSTM networks for sequence analysis
4. **Real-time Processing**: Optimized inference pipeline for live translation
"""
)
st.markdown(
"""
### 📊 Dataset Information
Our model is trained on the comprehensive [INCLUDE dataset](https://zenodo.org/records/4010759):
"""
)
# Dataset statistics table
dataset_stats = {
"Metric": [
"Categories", "Total Words", "Training Videos",
"Avg Videos/Class", "Avg Video Length", "Resolution", "Frame Rate"
],
"Value": [
"15", "263", "4,257", "16.3", "2.57s", "1920x1080", "25fps"
]
}
st.table(pd.DataFrame(dataset_stats))
# Display dataset processing visualization
try:
categories_image = np.array(Image.open('original_project/categories_processed.png'))
st.image(categories_image, caption="📈 Processed Categories Distribution")
except:
st.info("📊 Dataset visualization images will be displayed when available")
# Model architecture information
st.markdown(
"""
### 🧠 Neural Network Architecture
```python
# TechMatrix Solvers LSTM Translation Model
model = Sequential([
Input(shape=(20, 156)), # 20-frame temporal window
Masking(mask_value=0.),
BatchNormalization(),
Bidirectional(LSTM(32, recurrent_dropout=0.2, return_sequences=True)),
Dropout(0.2),
Bidirectional(LSTM(32, recurrent_dropout=0.2)),
Dense(32, activation='elu'),
BatchNormalization(),
Dropout(0.2),
Dense(len(expression_mapping), activation='softmax')
])
```
**Model Statistics:**
- Total Parameters: 82,679 (322.96 KB)
- Trainable Parameters: 82,239 (321.25 KB)
- Input Features: 156-dimensional vectors
- Temporal Window: 20 frames
"""
)
# Technology stack
col1, col2 = st.columns(2)
with col1:
st.markdown(
"""
### 🛠️ Technology Stack
**Frontend & UI:**
- Streamlit (Interactive Web App)
- Custom CSS Styling
- Responsive Design
**Deep Learning:**
- Keras/TensorFlow Backend
- PyTorch Integration
- LSTM Networks
- OpenPose Models
"""
)
with col2:
st.markdown(
"""
### 📱 Key Features
**Real-time Processing:**
- Live video analysis
- Pose keypoint extraction
- Temporal sequence modeling
- Confidence scoring
**User Experience:**
- Intuitive interface
- Visual feedback
- Progress tracking
- Result visualization
"""
)
# System Information
st.markdown("### 🔧 System Information")
col1, col2 = st.columns(2)
with col1:
st.write(f"**Python Version:** {platform.python_version()}")
st.write(f"**FFmpeg:** {shutil.which('ffmpeg') or 'Not found'}")
st.write(f"**FFprobe:** {shutil.which('ffprobe') or 'Not found'}")
with col2:
try:
st.write(f"**OpenCV Version:** {cv2.__version__}")
except:
st.write("**OpenCV:** Not available")
try:
import torch
st.write(f"**PyTorch:** {torch.__version__}")
st.write(f"**Keras:** {keras.__version__}")
except:
st.write("**PyTorch/Keras:** Not available")
# Team contact information
st.markdown(
"""
### 📞 Contact Information
**TechMatrix Solvers Team:**
| Name | Role | Email | Phone |
|------|------|-------|---------|
| **Abhay Gupta** | Team Lead | contact2abhaygupta6187@gmail.com | 8115814535 |
| **Kripanshu Gupta** | Backend Developer | guptakripanshu83@gmail.com | 7067058400 |
| **Dipanshu Patel** | UI/UX Designer | dipanshupatel43@gmail.com | 9294526404 |
| **Bhumika Patel** | Deployment & Presenter | bp7249951@gmail.com | 9302271422 |
**Institution:** Shri Ram Group of Institutions
### 📚 Documentation
For detailed technical documentation and implementation details, please refer to our
[comprehensive documentation](https://docs.google.com/document/d/1mzr2KGHRJT5heUjFF20NQ3Gb89urpjZJ/edit?usp=sharing).
---
**© 2024 TechMatrix Solvers - Innovating Accessible Technology Solutions**
"""
)
elif app_mode == 'Test Video Translation':
# Video selection interface
st.markdown("## 🎥 Test Video Translation")
# Load test data dynamically
with st.spinner("Loading test data..."):
try:
testing_df, test_files_df = load_test_data()
st.success("✅ Test data loaded successfully!")
except Exception as e:
st.error(f"❌ Failed to load test data: {e}")
st.stop()
category = st.sidebar.selectbox(
'Choose Category',
np.sort(test_files_df['Category'].unique(), axis=-1, kind='mergesort')
)
# Filter by category
category_mask = (test_files_df['Category'] == category)
test_files_category = test_files_df[category_mask]
class_name = st.sidebar.selectbox(
'Choose Class',
np.sort(test_files_category['Class'].unique(), axis=-1, kind='mergesort')
)
# Filter by class
class_mask = (test_files_df['Class'] == class_name)
filename = st.sidebar.selectbox(
'Choose File',
np.sort(test_files_category[class_mask]['Filename'].unique(), axis=-1, kind='mergesort')
)
# Display selection info
st.info(f"📂 Selected: {category} → {class_name} → {filename}")
if st.sidebar.button("🚀 Start Translation", type="primary"):
# Filter test data for selected video
data_mask = ((testing_df['FileName'] == filename) &
(testing_df['Type'] == category) &
(testing_df['Expression'] == class_name))
window_size = 20
current_test_data = testing_df[data_mask]
if current_test_data.empty:
st.error(f"⚠️ No matching data found for: {filename} | {category} | {class_name}")
st.stop()
else:
st.success(f"✅ Loaded {current_test_data.shape[0]} frames for processing")
# Create time series data
X_test_processed, y_test_processed = create_time_series_sequences(
current_test_data, feature_columns_processed, label_columns, window_size=window_size
)
X_test_processed = np.array(X_test_processed)
# Configure Streamlit display options
st.set_option('deprecation.showfileUploaderEncoding', False)
st.sidebar.markdown('---')
st.markdown(
"""
<style>
[data-testid="stSidebar"][aria-expanded="true"] > div:first-child {
width: 400px;
}
[data-testid="stSidebar"][aria-expanded="false"] > div:first-child {
width: 400px;
margin-left: -400px;
}
</style>
""",
unsafe_allow_html=True,
)
st.sidebar.markdown('---')
st.markdown('## 📊 Translation Results')
# Progress tracking container
progress_container = st.empty()
with progress_container.container():
progress_df = pd.DataFrame([['--', '--']],
columns=['Frames Processed', 'Detected Sign'])
progress_table = st.table(progress_df)
# Video display container
video_display = st.empty()
st.markdown("<hr/>", unsafe_allow_html=True)
frame_display = st.empty()
# Download test video
video_file_path = hf_hub_download(
repo_id="sunilsarolkar/isl-test-data",
filename=f'test/{category}/{class_name}/{filename}',
repo_type="dataset"
)
if not os.path.exists(video_file_path):
st.error(f"⚠️ Video file not found: {video_file_path}")
st.stop()
# Initialize video capture
video_capture = cv2.VideoCapture(video_file_path)
# Get video metadata
probe_result = probe_video_info(video_file_path)
video_info = json.loads(probe_result.json)
video_stream = [stream for stream in video_info["streams"] if stream["codec_type"] == "video"][0]
input_fps = video_stream["avg_frame_rate"]
input_pix_fmt = video_stream["pix_fmt"]
input_vcodec = video_stream["codec_name"]
format_name = video_info["format"]["format_name"].split(",")[0]
# Video properties
width = int(video_capture.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps_input = int(video_capture.get(cv2.CAP_PROP_FPS))
# Processing variables
total_frames = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
frame_buffer = []
# Output video configuration
output_file = f"/tmp/techmatrix_output_{uuid.uuid4().hex}.{format_name}"
video_writer = None
weighted_predictions = {}
frame_predictions = {} # Reset for each video session
frame_idx = 0
try:
# Process each frame
for _, frame_data in current_test_data.iterrows():
if not video_capture.isOpened():
st.error(f"❌ Could not open video: {video_file_path}")
break
if video_capture.isOpened():
ret, frame = video_capture.read()
if len(frame_buffer) < window_size:
# Initial frames - build up buffer
if utils is not None:
visualization_canvas = utils.render_stick_model(
frame,
eval(frame_data['bodypose_circles']),
eval(frame_data['bodypose_sticks']),
eval(frame_data['handpose_edges']),
eval(frame_data['handpose_peaks'])
)
else:
visualization_canvas = frame # Use original frame if utils not available
# Add prediction plots
if utils is not None:
canvas_with_predictions = utils.create_bar_plot_visualization(
visualization_canvas, {},
f'Building Buffer - Frame {frame_idx + 1} [No Predictions Yet]',
visualization_canvas
)
canvas_with_predictions = utils.create_bar_plot_visualization(
canvas_with_predictions, weighted_predictions,
f'Weighted Average - Frame {frame_idx + 1} [No Predictions Yet]',
visualization_canvas
)
canvas_with_predictions = utils.add_bottom_padding(
canvas_with_predictions, (255, 255, 255), 100
)
else:
canvas_with_predictions = visualization_canvas # Use base canvas if utils not available
# Initialize video writer
if video_writer is None:
input_framesize = canvas_with_predictions.shape[:2]
video_writer = VideoWriter(output_file, input_fps, input_framesize,
input_pix_fmt, input_vcodec)
video_writer.write_frame(canvas_with_predictions)
# Update progress display
with progress_container.container():
progress_df = pd.DataFrame(
[[f'{frame_idx + 1}/{current_test_data.shape[0]}',
'<Building 20-frame buffer>']],
columns=['Frames Processed', 'Detected Sign']
)
progress_table = st.table(progress_df)
frame_buffer.append(frame)
# Display current frame
with video_display.container():
st.image(canvas_with_predictions, channels='BGR', use_column_width=True)
else:
# Process with full buffer - make predictions
frame_buffer[:-1] = frame_buffer[1:]
frame_buffer[-1] = frame
# Load translation model
translation_model = load_translation_model()
# Check if model loaded successfully
sequence_idx = frame_idx - 20 # Define sequence_idx for both cases
if translation_model is None:
st.error("❌ Translation model failed to load. Cannot make predictions.")
# Use dummy predictions to keep the visualization working
current_predictions = {"model_not_available": 0.0}
top_3_signs = ["model_not_available"]
top_3_probabilities = [0.0]
else:
# Make prediction on current window
prediction_output = translation_model(
X_test_processed[sequence_idx].reshape(
1, X_test_processed[sequence_idx].shape[0],
X_test_processed[sequence_idx].shape[1]
)
)
# Handle both PyTorch and Keras/TensorFlow models
try:
# Try PyTorch tensor operations first
prediction_output = prediction_output[0].cpu().detach().numpy()
except AttributeError:
# If it's a Keras model, it already returns NumPy arrays
prediction_output = prediction_output[0]
# Get top predictions
top_prediction_idx = np.argmax(prediction_output)
top_3_indices = prediction_output.argsort()[-3:][::-1]
top_3_signs = [get_sign_label(i) for i in top_3_indices]
top_3_probabilities = prediction_output[top_3_indices]
# Current frame predictions
current_predictions = {}
for sign, prob in zip(top_3_signs, top_3_probabilities):
current_predictions[sign] = prob
# Update frame-wise predictions for weighted average
for sign, prob in zip(top_3_signs, top_3_probabilities):
if sign not in frame_predictions:
frame_predictions[sign] = []
frame_predictions[sign].append(prob)
# Calculate weighted averages
for sign in frame_predictions:
sign_predictions = frame_predictions[sign]
sign_weights = [len(sign_predictions) for _ in range(len(sign_predictions))]
weighted_predictions[sign] = calculate_weighted_average(
sign_predictions, sign_weights
)
# Sort predictions by confidence
sorted_predictions = dict(
sorted(weighted_predictions.items(), key=lambda item: item[1], reverse=True)
)
# Create visualization
if utils is not None:
visualization_canvas = utils.render_stick_model(
frame,
eval(frame_data['bodypose_circles']),
eval(frame_data['bodypose_sticks']),
eval(frame_data['handpose_edges']),
eval(frame_data['handpose_peaks'])
)
else:
visualization_canvas = frame # Use original frame if utils not available
# Add prediction visualizations
if utils is not None:
canvas_with_predictions = utils.create_bar_plot_visualization(
visualization_canvas, current_predictions,
f'Current Window Prediction (Frames {sequence_idx + 1}-{frame_idx + 1})',
visualization_canvas
)
canvas_with_predictions = utils.create_bar_plot_visualization(
canvas_with_predictions, weighted_predictions,
f'Cumulative Weighted Average - Frame {frame_idx + 1}',
visualization_canvas
)
canvas_with_predictions = utils.add_bottom_padding(
canvas_with_predictions, (255, 255, 255), 100
)
else:
canvas_with_predictions = visualization_canvas # Use base canvas if utils not available
video_writer.write_frame(canvas_with_predictions)
# Get best prediction for display
if weighted_predictions:
best_sign = max(weighted_predictions, key=weighted_predictions.get)
best_confidence = weighted_predictions[best_sign]
else:
best_sign = "no_predictions"
best_confidence = 0.0
# Update progress display
with progress_container.container():
progress_df = pd.DataFrame(
[[f'{frame_idx + 1}/{current_test_data.shape[0]}',
f'{best_sign} ({best_confidence * 100:.2f}%)']],
columns=['Frames Processed', 'Detected Sign']
)
progress_table = st.table(progress_df)
# Display current frame
with video_display.container():
st.image(canvas_with_predictions, channels='BGR', use_column_width=True)
frame_idx += 1
# Finalize video processing
st.success("✅ Video processing completed!")
with video_display.container():
if video_writer is not None:
video_writer.close()
with open(output_file, 'rb') as video_file:
output_video_bytes = video_file.read()
st.video(output_video_bytes)
st.info(f"💾 Processed video saved: {output_file}")
else:
st.warning("⚠️ No video output generated")
finally:
# Clean up resources
if 'video_capture' in locals() and video_capture is not None:
video_capture.release()
if 'video_writer' in locals() and video_writer is not None:
video_writer.close()
# Note: cv2.destroyAllWindows() removed for headless compatibility
# Footer
st.markdown(
"""
---
<div style="text-align: center; color: #666;">
<p><strong>TechMatrix Solvers</strong> | Shri Ram Group of Institutions</p>
<p>Innovating Accessible Technology Solutions for Everyone 🚀</p>
</div>
""",
unsafe_allow_html=True
) |