https://github.com/saravanabalagi/pspnet_tf2
PSPNet in Tensorflow 2 with pretrained weights for ADE20k, CityScapes and VOC2012
https://github.com/saravanabalagi/pspnet_tf2
ade20k cityscapes pretrained-models pspnet tensorflow-2 tf2 voc2012
Last synced: 14 days ago
JSON representation
PSPNet in Tensorflow 2 with pretrained weights for ADE20k, CityScapes and VOC2012
- Host: GitHub
- URL: https://github.com/saravanabalagi/pspnet_tf2
- Owner: saravanabalagi
- License: mit
- Created: 2020-03-30T21:04:43.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2020-04-14T23:57:23.000Z (about 5 years ago)
- Last Synced: 2025-02-23T07:06:53.211Z (2 months ago)
- Topics: ade20k, cityscapes, pretrained-models, pspnet, tensorflow-2, tf2, voc2012
- Language: Python
- Size: 9.09 MB
- Stars: 5
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# PSPNet Tensorflow 2
Keras Pyramid Scene Parsing Network ported to tensorflow 2 from keras/tf_1.13.
- Caffe implementation: [PSPNet](https://github.com/hszhao/PSPNet)
- Py35 Keras Tensorflow1.13 implementation: [PSPNet-Keras-tensorflow](https://github.com/Vladkryvoruchko/PSPNet-Keras-tensorflow)
## Dependencies
- Tensorflow 2 (tensorflow / tensorflow-gpu / tensorflow-cpu)
- OpenCV (opencv-python / opencv-contrib-python)
## Pretrained weights
Pretrained weights can be downloaded here:
| H5 Weights | Architecture | Numpy Weights |
|------------|--------------|---------------|
| [pspnet50_ade20k.h5](https://www.dropbox.com/s/7eyuzmag8df41j4/pspnet50_ade20k.h5?dl=0) | [pspnet50_ade20k.json](https://www.dropbox.com/s/xy7gs4g2def5z89/pspnet50_ade20k.json?dl=0) | [pspnet50_ade20k.npy](https://www.dropbox.com/s/z8la9ugpdss8k8q/pspnet50_ade20k.npy?dl=0) |
| [pspnet101_cityscapes.h5](https://www.dropbox.com/s/oymx9ktu6zrv7vz/pspnet101_cityscapes.h5?dl=0) | [pspnet101_cityscapes.json](https://www.dropbox.com/s/pofkdnf59nbs5w0/pspnet101_cityscapes.json?dl=0) | [pspnet101_cityscapes.npy](https://www.dropbox.com/s/2tdl01ihse7p9sr/pspnet101_cityscapes.npy?dl=0) |
| [pspnet101_voc2012.h5](https://www.dropbox.com/s/lqkmukeuo78cbcs/pspnet101_voc2012.h5?dl=0) | [pspnet101_voc2012.json](https://www.dropbox.com/s/i9f2p3q1d4wohd3/pspnet101_voc2012.json?dl=0) | [pspnet101_voc2012.npy](https://www.dropbox.com/s/yp4im80m72r6h98/pspnet101_voc2012.npy?dl=0) |
Download weights in
- `.h5` and `.json` format and place them at `weights/keras` or
- `.npy` and place them at `weights/npy`
Find example [notebook](save_and_load.ipynb) which demonstrates save and load.
## Usage:
```sh
# python pspnet.py -m -i -o [-other_arguments]
python pspnet.py -m pspnet101_cityscapes -i example_images/cityscapes.jpg -o example_results/cityscapes.jpg -s -ms -f
python pspnet.py -m pspnet101_voc2012 -i example_images/pascal_voc.jpg -o example_results/pascal_voc.jpg -s -ms -f
python pspnet.py -m pspnet50_ade20k -i example_images/ade20k.jpg -o example_results/ade20k.jpg -s -ms -f
```
List of arguments:
```sh
-m --model - which model to use: 'pspnet50_ade20k', 'pspnet101_cityscapes', 'pspnet101_voc2012'
--id - (int) GPU Device id. Default 0
-s --sliding - Use sliding window
-f --flip - Additional prediction of flipped image
-ms --multi_scale - Predict on multiscale images
```
 Batch Predict on GPU, check source [here](https://github.com/saravanabalagi/pspnet_tf2/blob/master/pspnet.py#L49)
## Keras results:
| Input | Segmented | Blended | Probe |
|-------|-----------|---------|-------|
|  |  |  |  |
|  |  |  |  |
|  |  |  |  |
## Implementation
* The interpolation layer is implemented as custom layer "Interp"
* Forward step takes about ~1 sec on single image
* Memory usage can be optimized with:
```python
# before calling any of the tf functions
for gpu in tf.config.experimental.list_physical_devices('GPU'):
tf.config.experimental.set_memory_growth(gpu, True)
# if you want to restrict total memory you can try
# tf.config.experimental.set_memory_growth(gpu, True)
```