Awesome

Unsupervised Domain Adaptive Re-Identification

Implementation of the paper Unsupervised Domain Adaptive Re-Identification: Theory and Practice.

The selftraining scheme proposed in the paper is simple yet effective.

Illustration of the selftraining scheme.

Running the experiments

Step 1: Train on source dataset

Run source_train.py via

python source_train.py \
    --dataset <name_of_source_dataset>\
    --resume <dir_of_source_trained_model>\
    --data_dir <dir_of_source_data>\
    --logs_dir <dir_to_save_source_trained_model>

To replicate the results in the paper, you can download pre-trained models on Market1501 from GoogleDrive and on DukeMTMC from GoogleDrive. Our models are trained with PyTorch 0.3.

Step 2: Run self-training

python selftraining.py \
    --src_dataset <name_of_source_dataset>\
    --tgt_dataset <name_of_target_dataset>\
    --resume <dir_of_source_trained_model>\
    --data_dir <dir_of_source_target_data>\
    --logs_dir <dir_to_save_model_after_adaptation>

Other code

dw_example.ipynb is the file for replicating Figure 6 in the paper.

Results

Step 1: After training on source dataset

Source Dataset	Rank-1	mAP
DukeMTMC	80.8	65.4
Market1501	91.6	78.2
CUHK03	48.79	46.95
MSMT17	69.82	42.48

Step 2: After adaptation

<table> <tr> <th rowspan="2">SRC --> TGT</th> <th colspan="2">Before Adaptation</th> <th colspan="2">After Adaptation</th> <th rowspan="2">Settings</th> </tr> <tr> <td>Rank-1</td> <td>mAP</td> <td>Rank-1</td> <td>mAP</td> </tr> <tr><td>CUHK --> Market</td><td>43.26</td><td>19.95</td><td>77.14</td><td>56.60</td><td>default</td></tr> <tr><td>CUHK --> DUKE</td><td>19.52</td><td>8.74</td><td>62.48</td><td>42.26</td><td>default</td></tr> <tr><td>CUHK --> MSMT</td><td>8.64</td><td>2.49</td><td>29.57</td><td>11.28</td><td>4GPU</td></tr> <tr><td>Market --> DUKE</td><td>27.3</td><td>11.9</td><td>68.4</td><td>49.0</td><td>default</td></tr> <tr><td>Market --> CUHK</td><td>4.07</td><td>4.53</td><td>20.32</td><td>20.85</td><td>default</td></tr> <tr><td>Market --> MSMT</td><td>8.37</td><td>2.54</td><td>30.54</td><td>12.04</td><td>4GPU, num_instances=8</td></tr> <tr><td>DUKE --> Market</td><td>46.8</td><td>19.1</td><td>75.8</td><td>53.7</td><td>default</td></tr> <tr><td>DUKE --> CUHK</td><td>4.43</td><td>4.56</td><td>9.89</td><td>10.32</td><td>default</td></tr> <tr><td>DUKE --> MSMT</td><td>12.38</td><td>3.82</td><td>39.22</td><td>15.99</td><td>4GPU, num_instances=8</td></tr> <tr><td>MSMT --> Market</td><td>49.47</td><td>23.71</td><td>80.94</td><td>59.97</td><td>4GPU</td></tr> <tr><td>MSMT --> DUKE</td><td>46.54</td><td>27.01</td><td>74.96</td><td>57.05</td><td>4GPU</td></tr> <tr><td>MSMT --> CUHK</td><td>10.71</td><td>11.59</td><td>16.21</td><td>16.56</td><td>4GPU</td></tr> </table>

FAQ

If you found difficulty in reproducing our reported results, please check the number of GPUs in your experiments. This may be caused by the un-sync BN layer implementation of PyTorch. Below we provide some training logs with the setting Duke --> Market, for helping you check your experiment settings.

With 1 GPU

Job ID	24312	24313	24314	24315	24316	24317
Rank 1	72.7	72.8	73.1	72.9	73.8	73.0

OneDrive Download Link

With 2 GPU

Job ID	24306	24307	24308	24309	24310	24311
Rank 1	75.1	75.7	77.5	76.1	77.5	77.3

OneDrive Download Link

Acknowledgement

Our code is based on open-reid.