Skip to content

Commit

Permalink
Update info for MaSTr1478
Browse files Browse the repository at this point in the history
  • Loading branch information
Lojze Žust committed Apr 13, 2023
1 parent cd3df7c commit 18955a4
Show file tree
Hide file tree
Showing 4 changed files with 16 additions and 9 deletions.
4 changes: 3 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -103,7 +103,7 @@ Currently available pretrained model weights. All models are evaluated on the MO
To train your own models, use the `train.py` script. For example, to reproduce the results of our experiments use the following steps:

1. Download and prepare the [MaSTr1325 dataset](https://box.vicos.si/borja/viamaro/index.html#mastr1325) (images and GT masks). Also download the context frames for the MaSTr1325 images [here](#data).
2. Edit the dataset configuration files (`configs/mastr_1325_train.yaml`, `configs/mastr1325_val.yaml`) so that they correctly point to the dataset directories.
2. Edit the dataset configuration files (`configs/mastr_1325_train.yaml`, `configs/mastr1325_val.yaml` and `configs/mastr153_all.yaml`) so that they correctly point to the dataset directories.
3. Use the `train.py` to train the network.

```bash
Expand All @@ -119,6 +119,8 @@ python train.py \

**Note:** Model training requires a large amount of GPU memory (>11 GB per GPU). If you use smaller GPUs, you can reduce the memory consumption by decreasing the number of backbone backpropagation steps (`--backbone-grad-steps`) or using a smaller context length (`--hist-len`).

**Note:** To reproduce training on MaSTr1478 use `--additional-train-config configs/mastr153_all.yaml` to specify the additional training examples.

### Logging and model weights

A log dir with the specified model name will be created inside the `output` directory. Model checkpoints and training logs will be stored here. At the end of the training the model weights are also exported to a `weights.pth` file inside this directory.
Expand Down
8 changes: 4 additions & 4 deletions configs/mastr1325_train.yaml
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
image_dir: images
image_hist_dir: images_hist
image_dir: /path/to/ds/images
image_hist_dir: /path/to/ds/images_hist
image_hist_len: 5
image_list: all_list_hist.txt
mask_dir: masks
image_list: /path/to/ds/all_list_hist.txt
mask_dir: /path/to/ds/masks
8 changes: 4 additions & 4 deletions configs/mastr1325_val.yaml
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
image_dir: images
image_hist_dir: images_hist
image_dir: /path/to/ds/images
image_hist_dir: /path/to/ds/images_hist
image_hist_len: 5
image_list: val_list_hist.txt
mask_dir: masks
image_list: /path/to/ds/val_list_hist.txt
mask_dir: /path/to/ds/masks
5 changes: 5 additions & 0 deletions configs/mastr153_all.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
image_dir: /path/to/ds/images
image_hist_dir: /path/to/ds/images_hist
image_hist_len: 5
image_list: /path/to/ds/all_list.txt
mask_dir: /path/to/ds/masks

0 comments on commit 18955a4

Please sign in to comment.