How brundha200/vision_ocr works
QwenVL Platform is a web app for document-image extraction, visual Q&A, audio transcription, and video-to-text processing. Users sign in, then submit media through the browser interface; Flask handles requests and returns results for display or export. Image tasks use the Qwen vision-language model, while video processing can combine Whisper speech transcription with sampled-frame OCR. The excerpts establish the backend model calls and one browser-to-backend request; some frontend request wiring and page contents were not sampled, so those links are omitted.
Main entry points: app.py Session access · qwen_model.py Qwen vision model · index.html Platform UI · script.js UI interactions
Components 12
Web interface
- Platform UI
index.html - UI interactions
script.js - Login page
login.html
Flask service
- Session access
app.py - Image extraction
app.py - Visual Q&A
app.py - Audio transcription
app.py - Video processing
app.py
Media inference
- Qwen vision model
qwen_model.py - Whisper handler
qwen_model.py
Other components
- User
- FFmpeg
Connections 10
- User → Platform UI: uses
- User → Session access: signs in
app.py - Session access → Login page: renders
app.py - Session access → Platform UI: renders
app.py - UI interactions → Session access: posts question
script.js - Image extraction → Qwen vision model: extracts text
app.py - Visual Q&A → Qwen vision model: asks image
app.py - Video processing → FFmpeg: extracts audio
app.py - Video processing → Whisper handler: transcribes speech
app.py - Video processing → Qwen vision model: reads sampled frames
app.py