Vision Models
Collection
Vision models for object detection and image recognition. Includes RT-DETR series (v1/v2/v3/v4) and other computer vision models. β’ 2 items β’ Updated
Official weights for RT-DETR (Real-Time DEtection TRansformer) series.
| Version | Directory | Files | Paper | Status |
|---|---|---|---|---|
| v1 | v1/ |
R18 / R50 / R101 (3) | CVPR 2024 | β Available |
| v2 | v2/ |
S / M / L / X (4) | 2024 | β Available |
| v3 | v3/ |
β | WACV 2025 | β PaddlePaddle only |
| v4 | v4/ |
S / M / L / X (4) | ECCV 2026 | β Available |
| Version | Model | Backbone | Size | APval |
|---|---|---|---|---|
| v1 | R18 | ResNet-18 | 77.2 MB | 46.4 |
| v1 | R50 | ResNet-50 | 164.4 MB | 53.1 |
| v1 | R101 | ResNet-101 | 293.3 MB | 54.3 |
| v2 | S | ResNet-18 | 77.4 MB | 48.1 |
| v2 | M | ResNet-50 | 140.5 MB | 51.9 |
| v2 | L | ResNet-50 | 164.5 MB | 53.4 |
| v2 | X | ResNet-101 | 293.5 MB | 54.3 |
| v4 | S | HGNetv2-S | 161.4 MB | 50.0 |
| v4 | M | HGNetv2-M | 303.2 MB | 52.5 |
| v4 | L | HGNetv2-L | 482.0 MB | 53.8 |
| v4 | X | HGNetv2-X | 963.7 MB | 54.7 |
File names are listed in each version's sub-directory README.
from huggingface_hub import hf_hub_download
# Download v4-S
hf_hub_download(
repo_id="eebyp/RT-DETR",
filename="v4/RTv4-S-hgnet.pth",
local_dir="./model"
)
# Download v2-S
hf_hub_download(
repo_id="eebyp/RT-DETR",
filename="v2/rtdetrv2_r18vd_120e_coco_rerun_48.1.pth",
local_dir="./model"
)
# Download v1-R50
hf_hub_download(
repo_id="eebyp/RT-DETR",
filename="v1/rtdetr_r50vd_6x_coco_from_paddle.pth",
local_dir="./model"
)
| Version | Year | Conference | Backbone | Key Contribution |
|---|---|---|---|---|
| v1 | 2024 | CVPR 2024 | ResNet | First real-time end-to-end detector |
| v2 | 2024 | β | ResNet | Improved baseline with bag-of-freebies |
| v3 | 2025 | WACV 2025 | ResNet | Hierarchical dense positive supervision (PaddlePaddle only) |
| v4 | 2026 | ECCV 2026 | HGNetv2 | Vision foundation model distillation |
Apache 2.0
Note: This repository is hosted for personal research purposes only. For production use, please refer to the official repositories above.