Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- object-detection
|
| 5 |
+
- instance-segmentation
|
| 6 |
+
- visdet
|
| 7 |
+
- pytorch
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# VisDet Model Weights
|
| 11 |
+
|
| 12 |
+
This repository contains pretrained weights for [VisDet](https://github.com/BinItAI/visdet),
|
| 13 |
+
a streamlined object detection and instance segmentation library.
|
| 14 |
+
|
| 15 |
+
## Models
|
| 16 |
+
|
| 17 |
+
The weights are organized by source:
|
| 18 |
+
- `openmmlab/`: Backbone weights from OpenMMLab's model zoo
|
| 19 |
+
- `mmcls/`: Classification backbone weights from MMClassification
|
| 20 |
+
|
| 21 |
+
## Usage
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
from visdet.apis import init_detector
|
| 25 |
+
|
| 26 |
+
# Use HuggingFace-hosted weights
|
| 27 |
+
model = init_detector(
|
| 28 |
+
config="configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco.py",
|
| 29 |
+
checkpoint="hf://GeorgePearse/visdet-weights/openmmlab/resnet50_msra-5891d200.pth"
|
| 30 |
+
)
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
Or in config files:
|
| 34 |
+
```python
|
| 35 |
+
model = dict(
|
| 36 |
+
backbone=dict(
|
| 37 |
+
init_cfg=dict(
|
| 38 |
+
type="Pretrained",
|
| 39 |
+
checkpoint="hf://GeorgePearse/visdet-weights/openmmlab/resnet50_msra-5891d200.pth"
|
| 40 |
+
)
|
| 41 |
+
)
|
| 42 |
+
)
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
## License
|
| 46 |
+
|
| 47 |
+
These weights are provided under the Apache 2.0 license, consistent with their original sources.
|