BiasLab2025/perception2

chore(forensics): finalize verification and trace validation\n\n- Fix unreachable background reasoning in inference.py\n- Enable enabling GPT reasoning by default in frontend\n- Verify correct data mapping for 15 naval fields\n- Clean up legacy code paths and ensure backward compatibility

feat(threat-assessment): implement naval threat analysis with GPT-4o\n\n- Rename utils/gpt_distance.py to utils/gpt_reasoning.py and update logic for 15 naval threat features\n- Add Pydantic schemas for NavalThreatAssessment in utils/schemas.py\n- Update backend (app.py, inference.py) to use new threat estimation and pass full metadata\n- refactor(frontend): render threat level badges and detailed feature table in UI

Fix: IndexError in ByteTracker

feat: replace SimpleTracker with ByteTrack

Show first frame radar when processed image appears

Add military-style radar with helicopter demo mode

Refine overlays: Hide on first frame and remove speed from labels

Respect CUDA_VISIBLE_DEVICES env var (do not delete it)

Fix NameError: import set_track_data in inference.py

Add writer loop error logging

Add aggressive worker debug logging and BaseException catch

Add granular logging for frame processing and ffmpeg

Fix backend worker pipeline stall on exception (add fallback fallback)

Fix video playback race condition, local file protocol, and backend worker crash

feat: Implement depth-based distance estimation and frontend integration

fix: Increase queue buffer to absorb GPT latency

feat: Radar historic traces and fix DETR crash

feat: Real-time radar directionality via track sync

fix: Import Full/Empty from queue to prevent NameError on backpressure

fix: Switch to gpt-4o-mini and verified queue safety

fix: Update static file path for deployment

feat: Continuous object tracking, speed estimation, and overlay syncing

Refactor frontend into modular ES modules in frontend/ folder

Fix tracking bugs: remove COCO fallback in HF mode, smooth velocity, and add track pruning

Use DETR no_timm revision to avoid timm dependency

Fix BGR to RGB conversion for DETR and GroundingDino inference

Fix inference concurrency bugs and enable default first-frame GPT

fix: Move CUDA_VISIBLE_DEVICES clearing to module import time

fix: Resolve IndentationError in DroneYoloDetector due to displaced import

feat: Implement SAHI Tiling for 4K video detection

optimize: Force high-res (1280px) inference for YOLO models to improve small object detection

perf: Tune batch sizes and queues for A10 GPUs

feat: Implement AsyncVideoReader for parallel video decoding

Simplify drone YOLO loading with ultralytics native HF Hub support

Uncomment CUDA_VISIBLE_DEVICES clearing to enable all GPUs

Update Lite model to use spencercdz/YOLOv8m_defence

Remove InternVL2 logic as requested

Fix internvl dependency subdirectory

Fix internvl dependency github url

Add internvl dependency and disable frontend caching

Fix SAM3 batch prediction shape mismatch and add InternVL2 to frontend

feat: Integrate InternVL2 and fix SAM3 segmentation batch size issue

Optimize streaming buffer and implement parallel depth pipeline

Refactor: Decouple depth estimator and suppress noisy logs

Implement Batch Inference & Queue Backpressure Fixes

fix: Handle empty GPT responses and log outputs

feat: Implement Tab 2 GPT-based ranging and remove legacy depth calc

Align Tab 2 Radar and UI with Tab 1

Enable throttled fallback tracking for Tab 2 consistency

Optimize MJPEG stream and fix first-frame UX