Commit 86251dd9 authored by Jong, Stefan1 de's avatar Jong, Stefan1 de
Browse files

Update README.md

parent b24bb6f2
......@@ -90,7 +90,7 @@ There are three possibilities to train TrackR-CNN on APPLE_MOTS:
- locate "load init" to your `models` folder, e.g. `"load_init": "/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/models/converted"`
- locate "KITTI_segtrack_data_dir" to your `train` folder e.g. `"/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/data/APPLE_MOTS/train/"`
- Set your batch size e.g. `"batch_size": 4`
- Set number of epoochs e.g. `"num_epochs": 4`
- Set number of epochs e.g. `"num_epochs": 4`
- run `python main.py configs/A_CONFIG_FILE`
2. Train APPLE_MOTS from KITTI MOTS checkpoint (fine-tuned on KITTI MOTS and pretrained on COCO and Mapillary)
......@@ -99,13 +99,13 @@ There are three possibilities to train TrackR-CNN on APPLE_MOTS:
- locate "load init" to your `models` folder, e.g. `"load_init": "/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/models/conv3d_sep2/conv3d_sep2-00000005"`
- locate "KITTI_segtrack_data_dir" to your `train` folder e.g. `"/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/data/APPLE_MOTS/train/"`
- Set your batch size e.g. `"batch_size": 4`
- Set number of epoochs e.g. `"num_epochs": 4`
- Set number of epochs e.g. `"num_epochs": 4`
- run `python main.py configs/A_CONFIG_FILE`
3. Train APPLE_MOTS on e.g. `3dconv8_24` by using more epochs (in this case 1 epoch more):
- Change in a config file:
- locate "load init" to your `models` folder, e.g. `"load_init": "/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/models/3dconv8_24/3dconv8_24-00000003"`
- Set number of epoochs `"num_epochs": 4` should be higher than 3dconv8_24-0000000**3** otherwise it doesn't train
- Set number of epochs `"num_epochs": 4` should be higher than 3dconv8_24-0000000**3** otherwise it doesn't train
- locate "KITTI_segtrack_data_dir" to your `train` folder e.g. `"/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/data/APPLE_MOTS/train/"`
- run `python main.py configs/A_CONFIG_FILE`
### Custom dataset
......@@ -117,7 +117,7 @@ There are three possibilities to train TrackR-CNN on APPLE_MOTS:
- Option 2 `"load_init": "/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/models/converted"`
- locate "KITTI_segtrack_data_dir" to your `train` folder e.g. `"/home/saidlab/Thesis_Stefan_de_Jong/rt-obj-tracking/algorithms/TrackR-CNN/data/APPLE_MOTS/train/"`
- Set your batch size e.g. `"batch_size": 4`
- Set number of epoochs e.g. `"num_epochs": 6`
- Set number of epochs e.g. `"num_epochs": 6`
- Set `"input_size_train"` and `"input_size_val"` to right resolutions.
- Calculate maximum number of detections in your dataset
- Go to home folder e.g. (`cd rt-obj-tracking`)
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment