Skip to content

Commit 2f6303d

Browse files
author
sunpeize
committed
add sparsercnn
1 parent 4eb9b85 commit 2f6303d

File tree

386 files changed

+49146
-0
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

386 files changed

+49146
-0
lines changed

‎.gitignore

+47
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,47 @@
1+
# output dir
2+
output
3+
instant_test_output
4+
inference_test_output
5+
6+
7+
*.png
8+
*.json
9+
*.diff
10+
11+
# compilation and distribution
12+
__pycache__
13+
_ext
14+
*.pyc
15+
*.pyd
16+
*.so
17+
detectron2.egg-info/
18+
build/
19+
dist/
20+
wheels/
21+
22+
# pytorch/python/numpy formats
23+
*.pth
24+
*.pkl
25+
*.npy
26+
27+
# ipython/jupyter notebooks
28+
*.ipynb
29+
**/.ipynb_checkpoints/
30+
31+
# Editor temporaries
32+
*.swn
33+
*.swo
34+
*.swp
35+
*~
36+
37+
# editor settings
38+
.idea
39+
.vscode
40+
_darcs
41+
42+
# project dirs
43+
/detectron2/model_zoo/configs
44+
/datasets/*
45+
!/datasets/*.*
46+
/projects/*/datasets
47+
/models

‎GETTING_STARTED.md

+79
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,79 @@
1+
## Getting Started with Detectron2
2+
3+
This document provides a brief intro of the usage of builtin command-line tools in detectron2.
4+
5+
For a tutorial that involves actual coding with the API,
6+
see our [Colab Notebook](https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5)
7+
which covers how to run inference with an
8+
existing model, and how to train a builtin model on a custom dataset.
9+
10+
For more advanced tutorials, refer to our [documentation](https://detectron2.readthedocs.io/tutorials/extend.html).
11+
12+
13+
### Inference Demo with Pre-trained Models
14+
15+
1. Pick a model and its config file from
16+
[model zoo](MODEL_ZOO.md),
17+
for example, `mask_rcnn_R_50_FPN_3x.yaml`.
18+
2. We provide `demo.py` that is able to run builtin standard models. Run it with:
19+
```
20+
cd demo/
21+
python demo.py --config-file ../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml \
22+
--input input1.jpg input2.jpg \
23+
[--other-options]
24+
--opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl
25+
```
26+
The configs are made for training, therefore we need to specify `MODEL.WEIGHTS` to a model from model zoo for evaluation.
27+
This command will run the inference and show visualizations in an OpenCV window.
28+
29+
For details of the command line arguments, see `demo.py -h` or look at its source code
30+
to understand its behavior. Some common arguments are:
31+
* To run __on your webcam__, replace `--input files` with `--webcam`.
32+
* To run __on a video__, replace `--input files` with `--video-input video.mp4`.
33+
* To run __on cpu__, add `MODEL.DEVICE cpu` after `--opts`.
34+
* To save outputs to a directory (for images) or a file (for webcam or video), use `--output`.
35+
36+
37+
### Training & Evaluation in Command Line
38+
39+
We provide a script in "tools/{,plain_}train_net.py", that is made to train
40+
all the configs provided in detectron2.
41+
You may want to use it as a reference to write your own training script.
42+
43+
To train a model with "train_net.py", first
44+
setup the corresponding datasets following
45+
[datasets/README.md](./datasets/README.md),
46+
then run:
47+
```
48+
cd tools/
49+
./train_net.py --num-gpus 8 \
50+
--config-file ../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml
51+
```
52+
53+
The configs are made for 8-GPU training.
54+
To train on 1 GPU, you may need to [change some parameters](https://arxiv.org/abs/1706.02677), e.g.:
55+
```
56+
./train_net.py \
57+
--config-file ../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml \
58+
--num-gpus 1 SOLVER.IMS_PER_BATCH 2 SOLVER.BASE_LR 0.0025
59+
```
60+
61+
For most models, CPU training is not supported.
62+
63+
To evaluate a model's performance, use
64+
```
65+
./train_net.py \
66+
--config-file ../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml \
67+
--eval-only MODEL.WEIGHTS /path/to/checkpoint_file
68+
```
69+
For more options, see `./train_net.py -h`.
70+
71+
### Use Detectron2 APIs in Your Code
72+
73+
See our [Colab Notebook](https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5)
74+
to learn how to use detectron2 APIs to:
75+
1. run inference with an existing model
76+
2. train a builtin model on a custom dataset
77+
78+
See [detectron2/projects](https://github.com/facebookresearch/detectron2/tree/master/projects)
79+
for more ways to build your project on detectron2.

‎INSTALL.md

+223
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,223 @@
1+
## Installation
2+
3+
Our [Colab Notebook](https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5)
4+
has step-by-step instructions that install detectron2.
5+
The [Dockerfile](docker)
6+
also installs detectron2 with a few simple commands.
7+
8+
### Requirements
9+
- Linux or macOS with Python ≥ 3.6
10+
- PyTorch ≥ 1.4
11+
- [torchvision](https://github.com/pytorch/vision/) that matches the PyTorch installation.
12+
You can install them together at [pytorch.org](https://pytorch.org) to make sure of this.
13+
- [pycocotools](https://github.com/cocodataset/cocoapi). Install it by `pip install pycocotools>=2.0.1`.
14+
- OpenCV, optional, needed by demo and visualization
15+
16+
17+
### Build Detectron2 from Source
18+
19+
gcc & g++ ≥ 5 are required. [ninja](https://ninja-build.org/) is recommended for faster build.
20+
After having them, run:
21+
```
22+
python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'
23+
# (add --user if you don't have permission)
24+
25+
# Or, to install it from a local clone:
26+
git clone https://github.com/facebookresearch/detectron2.git
27+
python -m pip install -e detectron2
28+
29+
# Or if you are on macOS
30+
CC=clang CXX=clang++ python -m pip install ......
31+
```
32+
33+
To __rebuild__ detectron2 that's built from a local clone, use `rm -rf build/ **/*.so` to clean the
34+
old build first. You often need to rebuild detectron2 after reinstalling PyTorch.
35+
36+
### Install Pre-Built Detectron2 (Linux only)
37+
38+
Choose from this table:
39+
40+
<table class="docutils"><tbody><th width="80"> CUDA </th><th valign="bottom" align="left" width="100">torch 1.5</th><th valign="bottom" align="left" width="100">torch 1.4</th> <tr><td align="left">10.2</td><td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
41+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu102/torch1.5/index.html
42+
</code></pre> </details> </td> <td align="left"> </td> </tr> <tr><td align="left">10.1</td><td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
43+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.5/index.html
44+
</code></pre> </details> </td> <td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
45+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.4/index.html
46+
</code></pre> </details> </td> </tr> <tr><td align="left">10.0</td><td align="left"> </td> <td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
47+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu100/torch1.4/index.html
48+
</code></pre> </details> </td> </tr> <tr><td align="left">9.2</td><td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
49+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu92/torch1.5/index.html
50+
</code></pre> </details> </td> <td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
51+
https://dl.fbaipublicfiles.com/detectron2/wheels/cu92/torch1.4/index.html
52+
</code></pre> </details> </td> </tr> <tr><td align="left">cpu</td><td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
53+
https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.5/index.html
54+
</code></pre> </details> </td> <td align="left"><details><summary> install </summary><pre><code>python -m pip install detectron2 -f \
55+
https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.4/index.html
56+
</code></pre> </details> </td> </tr></tbody></table>
57+
58+
59+
Note that:
60+
1. The pre-built package has to be used with corresponding version of CUDA and official PyTorch release.
61+
It will not work with a different version of PyTorch or a non-official build of PyTorch.
62+
2. Such installation is out-of-date w.r.t. master branch of detectron2. It may not be
63+
compatible with the master branch of a research project that uses detectron2 (e.g. those in
64+
[projects](projects) or [meshrcnn](https://github.com/facebookresearch/meshrcnn/)).
65+
66+
### Common Installation Issues
67+
68+
Click each issue for its solutions:
69+
70+
<details>
71+
<summary>
72+
Undefined symbols that contains TH,aten,torch,caffe2; missing torch dynamic libraries; segmentation fault immediately when using detectron2.
73+
</summary>
74+
<br/>
75+
76+
This usually happens when detectron2 or torchvision is not
77+
compiled with the version of PyTorch you're running.
78+
79+
If the error comes from a pre-built torchvision, uninstall torchvision and pytorch and reinstall them
80+
following [pytorch.org](http://pytorch.org). So the versions will match.
81+
82+
If the error comes from a pre-built detectron2, check [release notes](https://github.com/facebookresearch/detectron2/releases)
83+
to see the corresponding pytorch version required for each pre-built detectron2.
84+
Or uninstall and reinstall the correct pre-built detectron2.
85+
86+
If the error comes from detectron2 or torchvision that you built manually from source,
87+
remove files you built (`build/`, `**/*.so`) and rebuild it so it can pick up the version of pytorch currently in your environment.
88+
89+
If you cannot resolve this problem, please include the output of `gdb -ex "r" -ex "bt" -ex "quit" --args python -m detectron2.utils.collect_env`
90+
in your issue.
91+
</details>
92+
93+
<details>
94+
<summary>
95+
Undefined C++ symbols (e.g. `GLIBCXX`) or C++ symbols not found.
96+
</summary>
97+
<br/>
98+
Usually it's because the library is compiled with a newer C++ compiler but run with an old C++ runtime.
99+
100+
This often happens with old anaconda.
101+
Try `conda update libgcc`. Then rebuild detectron2.
102+
103+
The fundamental solution is to run the code with proper C++ runtime.
104+
One way is to use `LD_PRELOAD=/path/to/libstdc++.so`.
105+
106+
</details>
107+
108+
<details>
109+
<summary>
110+
"Not compiled with GPU support" or "Detectron2 CUDA Compiler: not available".
111+
</summary>
112+
<br/>
113+
CUDA is not found when building detectron2.
114+
You should make sure
115+
116+
```
117+
python -c 'import torch; from torch.utils.cpp_extension import CUDA_HOME; print(torch.cuda.is_available(), CUDA_HOME)'
118+
```
119+
120+
print valid outputs at the time you build detectron2.
121+
122+
Most models can run inference (but not training) without GPU support. To use CPUs, set `MODEL.DEVICE='cpu'` in the config.
123+
</details>
124+
125+
<details>
126+
<summary>
127+
"invalid device function" or "no kernel image is available for execution".
128+
</summary>
129+
<br/>
130+
Two possibilities:
131+
132+
* You build detectron2 with one version of CUDA but run it with a different version.
133+
134+
To check whether it is the case,
135+
use `python -m detectron2.utils.collect_env` to find out inconsistent CUDA versions.
136+
In the output of this command, you should expect "Detectron2 CUDA Compiler", "CUDA_HOME", "PyTorch built with - CUDA"
137+
to contain cuda libraries of the same version.
138+
139+
When they are inconsistent,
140+
you need to either install a different build of PyTorch (or build by yourself)
141+
to match your local CUDA installation, or install a different version of CUDA to match PyTorch.
142+
143+
* PyTorch/torchvision/Detectron2 is not built for the correct GPU architecture (aka. compute compatibility).
144+
145+
The compute compatibility included by PyTorch/detectron2/torchvision is available in the "architecture flags" in
146+
`python -m detectron2.utils.collect_env`. It must include
147+
the compute compatibility of your GPU, which can be found at [developer.nvidia.com/cuda-gpus](https://developer.nvidia.com/cuda-gpus).
148+
149+
If you're using pre-built PyTorch/detectron2/torchvision, they have included support for most popular GPUs already.
150+
If not supported, you need to build them from source.
151+
152+
When building detectron2/torchvision from source, they detect the GPU device and build for only the device.
153+
This means the compiled code may not work on a different GPU device.
154+
To recompile them for the correct compatiblity, remove all installed/compiled files,
155+
and rebuild them with the `TORCH_CUDA_ARCH_LIST` environment variable set properly.
156+
For example, `export TORCH_CUDA_ARCH_LIST=6.0,7.0` makes it compile for both P100s and V100s.
157+
</details>
158+
159+
<details>
160+
<summary>
161+
Undefined CUDA symbols; cannot open libcudart.so
162+
</summary>
163+
<br/>
164+
The version of NVCC you use to build detectron2 or torchvision does
165+
not match the version of CUDA you are running with.
166+
This often happens when using anaconda's CUDA runtime.
167+
168+
Use `python -m detectron2.utils.collect_env` to find out inconsistent CUDA versions.
169+
In the output of this command, you should expect "Detectron2 CUDA Compiler", "CUDA_HOME", "PyTorch built with - CUDA"
170+
to contain cuda libraries of the same version.
171+
172+
When they are inconsistent,
173+
you need to either install a different build of PyTorch (or build by yourself)
174+
to match your local CUDA installation, or install a different version of CUDA to match PyTorch.
175+
</details>
176+
177+
178+
<details>
179+
<summary>
180+
C++ compilation errors from NVCC
181+
</summary>
182+
183+
1. NVCC version has to match the CUDA version of your PyTorch.
184+
185+
2. NVCC has compatibility issues with certain versions of gcc. You sometimes need a different
186+
version of gcc. The version used by PyTorch can be found by `print(torch.__config__.show())`.
187+
</details>
188+
189+
190+
<details>
191+
<summary>
192+
"ImportError: cannot import name '_C'".
193+
</summary>
194+
<br/>
195+
Please build and install detectron2 following the instructions above.
196+
197+
Or, if you are running code from detectron2's root directory, `cd` to a different one.
198+
Otherwise you may not import the code that you installed.
199+
</details>
200+
201+
202+
<details>
203+
<summary>
204+
Any issue on windows.
205+
</summary>
206+
<br/>
207+
208+
Although detectron2 can be installed on windows with some effort (similar to [these](https://github.com/facebookresearch/pytorch3d/blob/master/INSTALL.md#2-install-from-a-local-clone)),
209+
we do not provide official support for it.
210+
211+
PRs that improves code compatibility on windows are welcome.
212+
</details>
213+
214+
<details>
215+
<summary>
216+
ONNX conversion segfault after some "TraceWarning".
217+
</summary>
218+
<br/>
219+
The ONNX package is compiled with a too old compiler.
220+
221+
Please build and install ONNX from its source code using a compiler
222+
whose version is closer to what's used by PyTorch (available in `torch.__config__.show()`).
223+
</details>

0 commit comments

Comments
 (0)