update readme

Author: GlassyWing

README.md

@@ -58,6 +58,10 @@ This library does not contain a feasible training program, please refer to the t
 
 [deepsort](https://github.com/ZQPei/deep_sort_pytorch)
 
+## Replacing Re-ID
+
+Please read the [tutorial](docs/how_to_replace_reid.md).
+
 ## References
 
 https://github.com/eriklindernoren/PyTorch-YOLOv3

docs/how_to_replace_reid.md

@@ -0,0 +1,43 @@
+# How to replace Re-ID model
+
+The repository use [fast-reid](https://github.com/JDAI-CV/fast-reid) as framework to support that.
+So you need custom and train your model with [fast-reid](https://github.com/JDAI-CV/fast-reid) first.
+
+Once you finish training your model with `fast-reid`,
+then you can replace your reid model as you do in fast-reid:
+
+1. Check out the branch `fast-reid`
+
+2. Put your custom re-id model defining in `deep_sort/deep/modeling/backbones`, 
+such as `resnet.py`, note that you must import your factory function in
+`deep_sort/deep/modeling/backbones/__init__.py`,
+E.g: `from .resnet import build_resnet_backbone`.
+see more [fast-reid](https://github.com/JDAI-CV/fast-reid).
+
+3. Put the config file generate by fast-reid in `config/deep`
+
+4. Config the DeepSort
+
+    There are serveral params that have to be setting: `config_file`, `MODEL.WEIGHTS`, `MODEL.DEVICE`, the config is need for `fast-reid`.
+
+    ```python
+
+    config = {"config_file": "config/deep/custom.yml",
+              "opts": ["MODEL.WEIGHTS",
+                       "weights/model_final.pth",
+                       "MODEL.DEVICE",
+                       "cuda:0"]}
+    cfg = setup_cfg(config)
+
+    tracker = DeepSort(cfg,
+                       min_confidence=1,
+                       nn_budget=30,
+                       n_init=3,
+                       max_iou_distance=0.7,
+                       max_dist=0.5,
+                       max_age=30)
+    ```
+
+5. Finished
+   
+   Now you can run the tracker with your Re-ID model.