Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- collaborative-perception
|
| 5 |
+
- v2x
|
| 6 |
+
- 3d-object-detection
|
| 7 |
+
- autonomous-driving
|
| 8 |
+
- lidar
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# FocalComm: Hard Instance-Aware Multi-Agent Perception (WACV 2026)
|
| 12 |
+
|
| 13 |
+
Pretrained checkpoints for [FocalComm](https://github.com/scdrand23/FocalComm), a collaborative perception framework that exchanges hard-instance-oriented features among connected agents via progressive Hard Instance Mining (HIM) and Query-guided Adaptive Feature Fusion (QAFF).
|
| 14 |
+
|
| 15 |
+
## Checkpoints
|
| 16 |
+
|
| 17 |
+
| File | Dataset | Config | AP@0.3/AP@0.5 |
|
| 18 |
+
|------|---------|--------|---------------|
|
| 19 |
+
| `focalcomm_v2xreal/net_epoch50.pth` | V2X-Real (vehicle-centric) | FocalComm 3-stage HIM, epoch 50 | Car 91.6/88.4, Pedestrian 53.2/26.1, Truck 50.4/47.2, Overall 65.1/53.9 |
|
| 20 |
+
| `focalcomm_dairv2x/net_epoch50.pth` | DAIR-V2X | FocalComm, epoch 50 | Vehicle 73.3/66.4 |
|
| 21 |
+
|
| 22 |
+
Each checkpoint directory includes the exact training `config.yaml`. The DAIR-V2X checkpoint reproduces the paper result exactly; the V2X-Real checkpoint is the final epoch of the paper's training run (within ~2 mAP of the paper's Table 1 row).
|
| 23 |
+
|
| 24 |
+
## Usage
|
| 25 |
+
|
| 26 |
+
```bash
|
| 27 |
+
pip install -U huggingface_hub
|
| 28 |
+
hf download scdrand23/FocalComm --local-dir checkpoints
|
| 29 |
+
|
| 30 |
+
python focalcomm/tools/inference.py --model_dir checkpoints/focalcomm_v2xreal --fusion_method intermediate
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
See the [GitHub repository](https://github.com/scdrand23/FocalComm) for installation and dataset preparation.
|
| 34 |
+
|
| 35 |
+
## Citation
|
| 36 |
+
|
| 37 |
+
```bibtex
|
| 38 |
+
@inproceedings{shenkut2026focalcomm,
|
| 39 |
+
title={FocalComm: Hard Instance-Aware Multi-Agent Perception},
|
| 40 |
+
author={Shenkut, Dereje and Bhagavatula, Vijayakumar},
|
| 41 |
+
booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
|
| 42 |
+
year={2026}
|
| 43 |
+
}
|
| 44 |
+
```
|