Redet

## A: How to use Redet

## 1. Login

IP shall be under the Campus Local Area Network

```sh
ssh redet@10.15.3.46
```

---

## 2. conda activate

Activate conda redet

```conda
conda activate redet
```

---

## 3. torch version

The torch version is 1.8.1

```sh
$ python
Python 3.7.16 (default, Jan 17 2023, 22:20:44)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.__version__
'1.8.1+cu111'
>>> torch.cuda.is_available()
True
>>>
```

---

## 4. Change directory to ReDet

```sh
cd Torch/ReDet/
```

output

```sh
(redet) redet@super6-B650M-DS3H:~/Torch/ReDet$ ls
compile.sh  data  demo_large_image.py  DOTA_devkit         INSTALL.md  mmdet           network.png  requirements.txt  TECHNICAL_DETAILS.md  work_dirs
configs     demo  docs                 GETTING_STARTED.md  LICENSE     mmdet.egg-info  README.md    setup.py          tools
```

---

## 5. Train

Train example code 
```sh
 python tools/train.py configs/ReDet/ReDet_re50_refpn_1x_dota15.py
```

Train Redet

> Ignorance of the mismatch, which are not effect to the epoch generation

>ref: https://github.com/csuhan/ReDet/issues/12

> the ReDet Repo Owner, 
csuhan: It is normal and will not affect the inference.
The key filter in state_dicts is automatically generated by the equivariant networks (backbone), so we do not incldue this keys in pretrained models.

---

```sh
(redet) redet@super6-B650M-DS3H:~/Torch/ReDet$ python tools/train.py configs/ReDet/ReDet_re50_refpn_1x_dota15.py
2023-11-03 12:32:03,019 - INFO - Distributed training: False
/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/e2cnn/nn/modules/r2_conv/basisexpansion_singleblock.py:80: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (Triggered internally at  /pytorch/aten/src/ATen/native/IndexingUtils.h:30.)
  full_mask[mask] = norms.to(torch.uint8)
2023-11-03 12:32:14,101 - INFO - load model from: work_dirs/ReResNet_pretrain/re_resnet50_c8_batch256-12933bc2.pth
2023-11-03 12:32:14,109 - WARNING - The model and loaded state dict do not match exactly

unexpected key in source state_dict: head.fc.weight, head.fc.bias

missing keys in source state_dict: layer3.4.conv2.filter, layer4.0.conv1.filter, layer1.0.conv2.filter, layer1.2.conv3.filter, layer3.4.conv1.filter, layer4.2.conv3.filter, layer2.1.conv2.filter, layer1.0.conv3.filter, layer4.0.downsample.0.filter, layer2.2.conv1.filter, layer2.0.conv2.filter, layer4.2.conv2.filter, layer2.0.conv1.filter, layer4.0.conv2.filter, layer4.1.conv3.filter, layer3.3.conv2.filter, layer2.3.conv1.filter, layer1.0.conv1.filter, layer3.3.conv3.filter, layer3.5.conv3.filter, layer3.1.conv2.filter, layer4.2.conv1.filter, layer3.5.conv2.filter, layer2.1.conv1.filter, layer1.0.downsample.0.filter, layer2.2.conv2.filter, layer3.5.conv1.filter, layer2.0.downsample.0.filter, layer4.0.conv3.filter, layer3.2.conv3.filter, layer3.0.conv3.filter, layer2.2.conv3.filter, layer1.1.conv2.filter, layer3.2.conv2.filter, layer2.3.conv2.filter, layer3.2.conv1.filter, layer3.0.downsample.0.filter, layer4.1.conv1.filter, layer3.1.conv1.filter, layer1.1.conv3.filter, layer4.1.conv2.filter, layer1.2.conv2.filter, layer3.1.conv3.filter, layer1.1.conv1.filter, layer2.3.conv3.filter, layer2.0.conv3.filter, layer3.4.conv3.filter, layer3.0.conv2.filter, layer2.1.conv3.filter, layer3.3.conv1.filter, layer1.2.conv1.filter, conv1.filter, layer3.0.conv1.filter

loading annotations into memory...
Done (t=2.17s)
creating index...
index created!
2023-11-03 12:32:18,820 - INFO - Start running, host: redet@super6-B650M-DS3H, work_dir: /home/redet/Torch/ReDet/work_dirs/ReDet_re50_refpn_1x_dota15
2023-11-03 12:32:18,820 - INFO - workflow: [('train', 1)], max: 12 epochs
/home/redet/Torch/ReDet/mmdet/core/anchor/anchor_target.py:112: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (Triggered internally at  /pytorch/aten/src/ATen/native/IndexingUtils.h:30.)
  anchors = flat_anchors[inside_flags, :]
/home/redet/Torch/ReDet/mmdet/core/anchor/anchor_target.py:185: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (Triggered internally at  /pytorch/aten/src/ATen/native/IndexingUtils.h:30.)
  ret[inds] = data
/home/redet/Torch/ReDet/mmdet/core/anchor/anchor_target.py:189: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (Triggered internally at  /pytorch/aten/src/ATen/native/IndexingUtils.h:30.)
  ret[inds, :] = data
/home/redet/Torch/ReDet/mmdet/core/bbox/transforms.py:56: UserWarning: This overload of addcmul is deprecated:
        addcmul(Tensor input, Number value, Tensor tensor1, Tensor tensor2, *, Tensor out)
Consider using one of the following signatures instead:
        addcmul(Tensor input, Tensor tensor1, Tensor tensor2, *, Number value, Tensor out) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:1005.)
  gx = torch.addcmul(px, 1, pw, dx)  # gx = px + pw * dx
/home/redet/Torch/ReDet/mmdet/models/rbbox_heads/rbbox_head.py:314: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (Triggered internally at  /pytorch/aten/src/ATen/native/IndexingUtils.h:30.)
  bboxes_list.append(bboxes[keep_inds])
2023-11-03 12:33:03,511 - INFO - Epoch [1][50/7260]     lr: 0.00399, eta: 21:29:37, time: 0.889, data_time: 0.128, memory: 5800, loss_rpn_cls: 0.4930, loss_rpn_bbox: 0.1228, s0.rbbox_loss_cls: 0.5461, s0.rbbox_acc: 90.7285, s0.rbbox_loss_bbox: 0.4393, s1.rbbox_loss_cls: 0.4526, s1.rbbox_acc: 92.4283, s1.rbbox_loss_bbox: 0.0817, loss: 2.1355
```

---

The Epoch of Redet shall be like this

```sh
2023-11-03 12:33:03,511 - INFO - Epoch [1][50/7260]     lr: 0.00399, eta: 21:29:37, time: 0.889, data_time: 0.128, memory: 5800, loss_rpn_cls: 0.4930, loss_rpn_bbox: 0.1228, s0.rbbox_loss_cls: 0.5461, s0.rbbox_acc: 90.7285, s0.rbbox_loss_bbox: 0.4393, s1.rbbox_loss_cls: 0.4526, s1.rbbox_acc: 92.4283, s1.rbbox_loss_bbox: 0.0817, loss: 2.1355
2023-11-03 12:33:38,971 - INFO - Epoch [1][100/7260]    lr: 0.00465, eta: 19:18:46, time: 0.709, data_time: 0.047, memory: 5800, loss_rpn_cls: 0.2873, loss_rpn_bbox: 0.0952, s0.rbbox_loss_cls: 0.4344, s0.rbbox_acc: 91.3613, s0.rbbox_loss_bbox: 0.4885, s1.rbbox_loss_cls: 0.2908, s1.rbbox_acc: 94.3583, s1.rbbox_loss_bbox: 0.2014, loss: 1.7974
2023-11-03 12:34:19,711 - INFO - Epoch [1][150/7260]    lr: 0.00532, eta: 19:25:45, time: 0.815, data_time: 0.095, memory: 5800, loss_rpn_cls: 0.3094, loss_rpn_bbox: 0.1077, s0.rbbox_loss_cls: 0.5346, s0.rbbox_acc: 88.6152, s0.rbbox_loss_bbox: 0.5817, s1.rbbox_loss_cls: 0.3879, s1.rbbox_acc: 91.6497, s1.rbbox_loss_bbox: 0.2355, loss: 2.1569
2023-11-03 12:35:01,170 - INFO - Epoch [1][200/7260]    lr: 0.00599, eta: 19:34:06, time: 0.829, data_time: 0.071, memory: 5800, loss_rpn_cls: 0.2239, loss_rpn_bbox: 0.1054, s0.rbbox_loss_cls: 0.5205, s0.rbbox_acc: 88.5918, s0.rbbox_loss_bbox: 0.5066, s1.rbbox_loss_cls: 0.4100, s1.rbbox_acc: 90.6026, s1.rbbox_loss_bbox: 0.3745, loss: 2.1408
2023-11-03 12:35:44,329 - INFO - Epoch [1][250/7260]    lr: 0.00665, eta: 19:48:38, time: 0.863, data_time: 0.073, memory: 5800, loss_rpn_cls: 0.1942, loss_rpn_bbox: 0.0948, s0.rbbox_loss_cls: 0.5406, s0.rbbox_acc: 87.6523, s0.rbbox_loss_bbox: 0.5291, s1.rbbox_loss_cls: 0.3972, s1.rbbox_acc: 90.3339, s1.rbbox_loss_bbox: 0.4407, loss: 2.1965
2023-11-03 12:36:27,110 - INFO - Epoch [1][300/7260]    lr: 0.00732, eta: 19:56:20, time: 0.856, data_time: 0.084, memory: 5800, loss_rpn_cls: 0.2314, loss_rpn_bbox: 0.1064, s0.rbbox_loss_cls: 0.5227, s0.rbbox_acc: 87.3770, s0.rbbox_loss_bbox: 0.4589, s1.rbbox_loss_cls: 0.4190, s1.rbbox_acc: 89.8363, s1.rbbox_loss_bbox: 0.4143, loss: 2.1525
2023-11-03 12:37:09,011 - INFO - Epoch [1][350/7260]    lr: 0.00799, eta: 19:57:59, time: 0.838, data_time: 0.070, memory: 5800, loss_rpn_cls: 0.1731, loss_rpn_bbox: 0.0976, s0.rbbox_loss_cls: 0.4558, s0.rbbox_acc: 88.8516, s0.rbbox_loss_bbox: 0.4244, s1.rbbox_loss_cls: 0.3701, s1.rbbox_acc: 90.7472, s1.rbbox_loss_bbox: 0.4366, loss: 1.9577
2023-11-03 12:37:49,944 - INFO - Epoch [1][400/7260]    lr: 0.00865, eta: 19:55:32, time: 0.819, data_time: 0.063, memory: 5800, loss_rpn_cls: 0.1806, loss_rpn_bbox: 0.0867, s0.rbbox_loss_cls: 0.4761, s0.rbbox_acc: 87.5254, s0.rbbox_loss_bbox: 0.5364, s1.rbbox_loss_cls: 0.3642, s1.rbbox_acc: 89.9981, s1.rbbox_loss_bbox: 0.5080, loss: 2.1520
2023-11-03 12:38:31,804 - INFO - Epoch [1][450/7260]    lr: 0.00932, eta: 19:56:27, time: 0.837, data_time: 0.073, memory: 5800, loss_rpn_cls: 0.1668, loss_rpn_bbox: 0.0919, s0.rbbox_loss_cls: 0.5685, s0.rbbox_acc: 86.2832, s0.rbbox_loss_bbox: 0.6825, s1.rbbox_loss_cls: 0.3852, s1.rbbox_acc: 89.5976, s1.rbbox_loss_bbox: 0.5760, loss: 2.4708
2023-11-03 12:39:12,102 - INFO - Epoch [1][500/7260]    lr: 0.00999, eta: 19:52:32, time: 0.806, data_time: 0.063, memory: 5800, loss_rpn_cls: 0.1441, loss_rpn_bbox: 0.0866, s0.rbbox_loss_cls: 0.4604, s0.rbbox_acc: 88.2207, s0.rbbox_loss_bbox: 0.4981, s1.rbbox_loss_cls: 0.3332, s1.rbbox_acc: 90.9321, s1.rbbox_loss_bbox: 0.4650, loss: 1.9874
2023-11-03 12:39:57,781 - INFO - Epoch [1][550/7260]    lr: 0.01000, eta: 20:03:20, time: 0.914, data_time: 0.126, memory: 5800, loss_rpn_cls: 0.1727, loss_rpn_bbox: 0.1301, s0.rbbox_loss_cls: 0.5019, s0.rbbox_acc: 85.5586, s0.rbbox_loss_bbox: 0.5222, s1.rbbox_loss_cls: 0.3563, s1.rbbox_acc: 89.4112, s1.rbbox_loss_bbox: 0.5068, loss: 2.1900
2023-11-03 12:40:38,394 - INFO - Epoch [1][600/7260]    lr: 0.01000, eta: 20:00:01, time: 0.812, data_time: 0.068, memory: 5800, loss_rpn_cls: 0.1568, loss_rpn_bbox: 0.0970, s0.rbbox_loss_cls: 0.3941, s0.rbbox_acc: 89.5645, s0.rbbox_loss_bbox: 0.4318, s1.rbbox_loss_cls: 0.3255, s1.rbbox_acc: 91.3245, s1.rbbox_loss_bbox: 0.4171, loss: 1.8223
2023-11-03 12:41:20,589 - INFO - Epoch [1][650/7260]    lr: 0.01000, eta: 20:00:37, time: 0.844, data_time: 0.081, memory: 5800, loss_rpn_cls: 0.1481, loss_rpn_bbox: 0.0916, s0.rbbox_loss_cls: 0.4268, s0.rbbox_acc: 87.8652, s0.rbbox_loss_bbox: 0.4908, s1.rbbox_loss_cls: 0.3338, s1.rbbox_acc: 90.1052, s1.rbbox_loss_bbox: 0.5118, loss: 2.0028
2023-11-03 12:42:02,391 - INFO - Epoch [1][700/7260]    lr: 0.01000, eta: 20:00:14, time: 0.836, data_time: 0.073, memory: 5800, loss_rpn_cls: 0.1370, loss_rpn_bbox: 0.0931, s0.rbbox_loss_cls: 0.4624, s0.rbbox_acc: 86.5039, s0.rbbox_loss_bbox: 0.5225, s1.rbbox_loss_cls: 0.3469, s1.rbbox_acc: 89.8299, s1.rbbox_loss_bbox: 0.5761, loss: 2.1379
2023-11-03 12:42:43,537 - INFO - Epoch [1][750/7260]    lr: 0.01000, eta: 19:58:32, time: 0.823, data_time: 0.067, memory: 5800, loss_rpn_cls: 0.1212, loss_rpn_bbox: 0.0868, s0.rbbox_loss_cls: 0.4085, s0.rbbox_acc: 88.8770, s0.rbbox_loss_bbox: 0.4338, s1.rbbox_loss_cls: 0.3346, s1.rbbox_acc: 90.5601, s1.rbbox_loss_bbox: 0.5226, loss: 1.9076
2023-11-03 12:43:30,859 - INFO - Epoch [1][800/7260]    lr: 0.01000, eta: 20:08:05, time: 0.946, data_time: 0.147, memory: 5800, loss_rpn_cls: 0.1433, loss_rpn_bbox: 0.0933, s0.rbbox_loss_cls: 0.4642, s0.rbbox_acc: 86.4082, s0.rbbox_loss_bbox: 0.4624, s1.rbbox_loss_cls: 0.3748, s1.rbbox_acc: 89.2309, s1.rbbox_loss_bbox: 0.5453, loss: 2.0833
2023-11-03 12:44:11,527 - INFO - Epoch [1][850/7260]    lr: 0.01000, eta: 20:05:09, time: 0.813, data_time: 0.067, memory: 5800, loss_rpn_cls: 0.1242, loss_rpn_bbox: 0.0773, s0.rbbox_loss_cls: 0.4273, s0.rbbox_acc: 88.0254, s0.rbbox_loss_bbox: 0.4755, s1.rbbox_loss_cls: 0.3518, s1.rbbox_acc: 89.8915, s1.rbbox_loss_bbox: 0.5321, loss: 1.9881
2023-11-03 12:44:56,037 - INFO - Epoch [1][900/7260]    lr: 0.01000, eta: 20:08:36, time: 0.890, data_time: 0.115, memory: 5800, loss_rpn_cls: 0.1482, loss_rpn_bbox: 0.0833, s0.rbbox_loss_cls: 0.4152, s0.rbbox_acc: 88.2148, s0.rbbox_loss_bbox: 0.5131, s1.rbbox_loss_cls: 0.3478, s1.rbbox_acc: 90.0385, s1.rbbox_loss_bbox: 0.5006, loss: 2.0081
2023-11-03 12:45:41,526 - INFO - Epoch [1][950/7260]    lr: 0.01000, eta: 20:13:06, time: 0.910, data_time: 0.119, memory: 5800, loss_rpn_cls: 0.1366, loss_rpn_bbox: 0.0846, s0.rbbox_loss_cls: 0.4239, s0.rbbox_acc: 87.7871, s0.rbbox_loss_bbox: 0.5363, s1.rbbox_loss_cls: 0.3346, s1.rbbox_acc: 90.6918, s1.rbbox_loss_bbox: 0.4724, loss: 1.9884
2023-11-03 12:46:18,346 - INFO - Epoch [1][1000/7260]   lr: 0.01000, eta: 20:04:37, time: 0.736, data_time: 0.045, memory: 5800, loss_rpn_cls: 0.1110, loss_rpn_bbox: 0.0842, s0.rbbox_loss_cls: 0.3673, s0.rbbox_acc: 89.7480, s0.rbbox_loss_bbox: 0.5102, s1.rbbox_loss_cls: 0.2810, s1.rbbox_acc: 91.9225, s1.rbbox_loss_bbox: 0.4635, loss: 1.8172
2023-11-03 12:46:53,848 - INFO - Epoch [1][1050/7260]   lr: 0.01000, eta: 19:55:06, time: 0.710, data_time: 0.052, memory: 5800, loss_rpn_cls: 0.1025, loss_rpn_bbox: 0.0780, s0.rbbox_loss_cls: 0.3746, s0.rbbox_acc: 88.9590, s0.rbbox_loss_bbox: 0.4415, s1.rbbox_loss_cls: 0.2895, s1.rbbox_acc: 91.2423, s1.rbbox_loss_bbox: 0.4991, loss: 1.7853
2023-11-03 12:47:33,727 - INFO - Epoch [1][1100/7260]   lr: 0.01000, eta: 19:52:05, time: 0.798, data_time: 0.060, memory: 5800, loss_rpn_cls: 0.1007, loss_rpn_bbox: 0.0808, s0.rbbox_loss_cls: 0.3434, s0.rbbox_acc: 90.3027, s0.rbbox_loss_bbox: 0.3672, s1.rbbox_loss_cls: 0.2711, s1.rbbox_acc: 92.2888, s1.rbbox_loss_bbox: 0.3776, loss: 1.5409
```

---

## B: Why deploy Redet

## 1. linux add user redet
```sh
sudo adduser redet
sudo usermod redet -aG sudo
groups redet
```

Expected Output
```sh
redet : redet sudo
```

---

## 2. Login to redet

CLI switch to user redet

```sh
su - redet
```

## OR

IP shall be under the Campus Local Area Network

```sh
ssh redet@10.15.3.46
```

---

## 3. Install miniconda

miniconda is simplified anaconda, which is excellent and light-weighted

```sh
mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
```

initialize for bash and zsh shells

```sh
~/miniconda3/bin/conda init bash
~/miniconda3/bin/conda init zsh
```

ref: https://docs.conda.io/projects/miniconda/en/latest/

---

## 4. Exit and re-login to init the conda

Exit

```sh
exit
```

Re-login 
```sh
ssh redet@10.15.3.46
```

Expected Output 
with a '(base)' ahead

```sh
(base) redet@super6-B650M-DS3H:/home$
```

---

## 5. Create conda environment

Python Version 3.7

```sh
conda create -n redet python=3.7 -y
source activate redet
conda install cython
```

Expected Output

conda 
```sh
# To activate this environment, use
#     $ conda activate redet
```

cython

```sh
The following NEW packages will be INSTALLED:
  cython             pkgs/main/linux-64::cython-0.29.33-py37h6a678d5_0
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
```

---

## 6. Install mmcv

The lower mmcv may have conficts, so choose mmcv == 0.2.13

```sh
pip install mmcv==0.2.13
```

Expected Output
sometimes the output may carried with some relays of the mmcv

```sh
Successfully installed mmcv-0.2.13
```

---

## 7. Install Pytorch

Do not give a shit to the repo owner opinion due to nowadays' gigachad newest GPU did not fucking supoort the lower version pytorch 1.1/1.3

```sh
pip install torch==1.8.1+cu111 torchvision==0.9.1+cu111 torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html
```

> ref:  https://pytorch.org/get-started/previous-versions/

> BTW, the AT_CHECK (Pytorch < 1.5) iissue can be prefectly tackled by one prefect line in the later

Expected Output

```sh
Installing collected packages: typing-extensions, pillow, torch, torchvision, torchaudio
Successfully installed pillow-9.5.0 torch-1.8.1+cu111 torchaudio-0.8.1 torchvision-0.9.1+cu111 typing-extensions-4.7.1
```

---

##  8. Clone Redet

Clone the ReDet Repository

```sh
git clone https://github.com/csuhan/ReDet.git
cd ReDet
ls
```

Expected output
```sh
~/ReDet$ ls
compile.sh  demo_large_image.py  GETTING_STARTED.md  mmdet        requirements.txt      tools
configs     docs                 INSTALL.md          network.png  setup.py
demo        DOTA_devkit          LICENSE             README.md    TECHNICAL_DETAILS.md
```

---

## 9. Replace AT_CHECK with TORCH_CHECK

I would like to say generally it was the geneius perfect code which replace the ReDet project mmdet all `AT_CHECK` to `TORCH_CHECK`

```sh
find . -type f -exec sed -i 's/AT_CHECK/TORCH_CHECK/g' {} +
```

After that, the AT_CHECK issue tackled and then we can compile

ref: https://github.com/csuhan/ReDet/issues/1
ref: https://github.com/csuhan/ReDet/issues/1#issuecomment-907960207
ref: https://github.com/pytorch/pytorch/issues/36581

---

## 10. bash compile.sh

```sh
bash compile.sh
```

Expected Output without FAIL

```sh
~/ReDet$ bash compile.sh
Building roi align op...
running build_ext
building 'roi_align_cuda' extension
creating /home/redet/ReDet/mmdet/ops/roi_align/build
creating /home/redet/ReDet/mmdet/ops/roi_align/build/temp.linux-x86_64-cpython-37
creating /home/redet/ReDet/mmdet/ops/roi_align/build/temp.linux-x86_64-cpython-37/src
Emitting ninja build file /home/redet/ReDet/mmdet/ops/roi_align/build/temp.linux-x86_64-cpython-37/build.ninja...
Compiling objects...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
[1/2] /usr/local/cuda-11.3/bin/nvcc --generate-dependencies-with-compile --dependency-output /home/redet/ReDet/mmdet/ops/roi_align/build/temp.linux-x86_64-cpython-37/src/roi_align_kernel.o.d -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/TH -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/THC -I/usr/local/cuda-11.3/include -I/home/redet/miniconda3/envs/redet/include/python3.7m -c -c /home/redet/ReDet/mmdet/ops/roi_align/src/roi_align_kernel.cu -o /home/redet/ReDet/mmdet/ops/roi_align/build/temp.linux-x86_64-cpython-37/src/roi_align_kernel.o -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options ''"'"'-fPIC'"'"'' -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1011"' -DTORCH_EXTENSION_NAME=roi_align_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -gencode=arch=compute_86,code=compute_86 -gencode=arch=compute_86,code=sm_86 -std=c++14
/home/redet/ReDet/mmdet/ops/roi_align/src/roi_align_kernel.cu: In lambda function:
/home/redet/ReDet/mmdet/ops/roi_align/src/roi_align_kernel.cu:133:46: warning: ‘at::DeprecatedTypeProperties& at::Tensor::type() const’ is deprecated: Tensor.type() is deprecated. Instead use Tensor.options(), which in many cases (e.g. in a constructor) is a drop-in replacement. If you were using data from type(), that is now available from Tensor itself, so instead of tensor.type().scalar_type(), use tensor.scalar_type() instead and instead of tensor.type().backend() use tensor.device(). [-Wdeprecated-declarations]
   AT_DISPATCH_FLOATING_TYPES_AND_HALF(
                                              ^                                                                     
```

---

Compilation [1/2]

```sh
copying build/lib.linux-x86_64-cpython-37/masked_conv2d_cuda.cpython-37m-x86_64-linux-gnu.so ->
Building poly_nms op...
running build_ext
building 'poly_nms_cuda' extension
creating /home/redet/ReDet/mmdet/ops/poly_nms/build
creating /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37
creating /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/src
Emitting ninja build file /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/build.ninja...
Compiling objects...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
[1/2] /usr/local/cuda-11.3/bin/nvcc --generate-dependencies-with-compile --dependency-output /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/src/poly_nms_kernel.o.d -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/TH -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/THC -I/usr/local/cuda-11.3/include -I/home/redet/miniconda3/envs/redet/include/python3.7m -c -c /home/redet/ReDet/mmdet/ops/poly_nms/src/poly_nms_kernel.cu -o /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/src/poly_nms_kernel.o -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options ''"'"'-fPIC'"'"'' -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1011"' -DTORCH_EXTENSION_NAME=poly_nms_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -gencode=arch=compute_86,code=compute_86 -gencode=arch=compute_86,code=sm_86 -std=c++14
```

---

Compilation [2/2]

```sh
[2/2] c++ -MMD -MF /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/src/poly_nms_cuda.o.d -pthread -B /home/redet/miniconda3/envs/redet/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/TH -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/torch/include/THC -I/usr/local/cuda-11.3/include -I/home/redet/miniconda3/envs/redet/include/python3.7m -c -c /home/redet/ReDet/mmdet/ops/poly_nms/src/poly_nms_cuda.cpp -o /home/redet/ReDet/mmdet/ops/poly_nms/build/temp.linux-x86_64-cpython-37/src/poly_nms_cuda.o -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1011"' -DTORCH_EXTENSION_NAME=poly_nms_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++14
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
```

The Comliation tail

```sh
copying build/lib.linux-x86_64-cpython-37/poly_nms_cuda.cpython-37m-x86_64-linux-gnu.so ->
Building cpu_nms...
running build_ext
skipping 'bbox.c' Cython extension (up-to-date)
building 'bbox' extension
creating build
creating build/temp.linux-x86_64-cpython-37
gcc -pthread -B /home/redet/miniconda3/envs/redet/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/redet/miniconda3/envs/redet/lib/python3.7/site-packages/numpy/core/include -I/home/redet/miniconda3/envs/redet/include/python3.7m -c bbox.c -o build/temp.linux-x86_64-cpython-37/bbox.o -Wno-cpp -Wno-unused-function
gcc -pthread -shared -B /home/redet/miniconda3/envs/redet/compiler_compat -L/home/redet/miniconda3/envs/redet/lib -Wl,-rpath=/home/redet/miniconda3/envs/redet/lib -Wl,--no-as-needed -Wl,--sysroot=/ build/temp.linux-x86_64-cpython-37/bbox.o -o /home/redet/ReDet/mmdet/core/bbox/bbox.cpython-37m-x86_64-linux-gnu.so
```

---

## 11. Pip install packages

```sh
pip install -e .
python setup.py develop
```

Expected Output

```sh
Successfully installed cycler-0.11.0 e2cnn-0.2.3 fonttools-4.38.0 kiwisolver-1.4.5 matplotlib-3.5.3 mmdet-0.6.0+f4a8e7d mpmath-1.3.0 packaging-23.2 pillow-6.2.2 pycocotools-2.0.7 pyparsing-3.1.1 python-dateutil-2.8.2 scipy-1.7.3 shapely-2.0.2 sympy-1.10.1 terminaltables-3.1.10 tqdm-4.66.1
```

---

## 12. Install DOTA_devkit

The ReDet repot has the devkit

```sh
sudo apt-get install swig
cd DOTA_devkit
swig -c++ -python polyiou.i
python setup.py build_ext --inplace
```

---

## 13. Data preparation

The Dota dataset can be downloaded online

> https://captain-whu.github.io/DOTA/dataset.html

And it the DOTA datasets shall be like this coco format with json

These two GETTING_STARTED.md are very similar

> https://github.com/csuhan/ReDet/blob/master/GETTING_STARTED.md

> https://github.com/dingjiansw101/AerialDetection/blob/master/GETTING_STARTED.md

---

```sh
data
├── dota15
│   ├── test
│   │   └── images
│   ├── train
│   │   ├── images
│   │   └── labelTxt
│   └── val
│       ├── images
│       └── labelTxt
├── dota15_1024
│   ├── test1024
│   │   ├── DOTA1_5_test1024.json
│   │   └── images
│   └── trainval1024
│       ├── DOTA1_5_trainval1024.json
│       ├── images
│       └── labelTxt
└── test_info.json
```

---

## 14. work_dirs pth preparation

the work_dirs pth can be downloaded from the ReDet Repo

> https://github.com/csuhan/ReDet#benchmark-and-model-zoo

```sh
work_dirs/
├── ReDet_re101_refpn_1x_dota15-2c64af7b.pth
├── ReDet_re50_refpn_1x_dota15
│   └── ReDet_re50_refpn_1x_dota15-7f2d6dda.pth
├── ReDet_re50_refpn_1x_dota15_ms-9d1a523c.pth
├── ReDet_re50_refpn_1x_dota1-a025e6b1.pth
├── ReDet_re50_refpn_1x_dota1_ms-ba948633.pth
├── ReDet_re50_refpn_3x_hrsc2016
│   └── ReDet_re50_refpn_3x_hrsc2016-d1b4bd29.pth
├── re_resnet50_c8_batch256
│   └── epoch_100.pth
└── ReResNet_pretrain
    ├── re_resnet50_c8_batch256-12933bc2.pth
    └── re_resnet50_c8_batch256-25b16846.pth
```

---

## 15. Train the redet

Train example code 
```sh
 python tools/train.py configs/ReDet/ReDet_re50_refpn_1x_dota15.py
```

Train Redet

> Ignorance of the mismatch, which are not effect to the epoch generation

>ref: https://github.com/csuhan/ReDet/issues/12

---

### The directory prepare for training

```sh
ReDet
├── compile.sh
├── configs
├── demo
├── demo_large_image.py
├── docs
├── DOTA_devkit
├── GETTING_STARTED.md
├── INSTALL.md
├── LICENSE
├── mmdet
├── mmdet.egg-info
├── network.png
├── README.md
├── requirements.txt
├── setup.py
├── TECHNICAL_DETAILS.md
└── tools
```

---

unexpected key in source state_dict: head.fc.weight, head.fc.bias

---

The Epoch of Redet shall be like this