Running UPSNet on google colab. Following is the description:
- UPSNet_Colab_GDrive_DataSet.ipynb - To test and train the UPSNet, keeping cityscapes dataset on GDrive.
- UPSNet_Colab_Local_DataSet.ipynb - To test and train the UPSNet, keeping cityscapes dataset in colab space.
- TrainedModel - Contains trained model files, which can be used directly for testing purposes. These files need to be placed in the output folder at the location: output\upsnet\cityscapes\upsnet_resnet50_cityscapes_4gpu\train.
- output - This folder contains the output of the test.
- Changed_Files - This folder included the changed files; required to train and test the UPSNet using the above-mentioned notebooks.
Train: For training space on google colab is less and also colab's session last for 8-12 hour; so we need to save the snapshot and next time start training from this snapshot onwards. Test: For directly testing upload this output folder in UPSNet code and remove val directory from output\upsnet\cityscapes\upsnet_resnet50_cityscapes_4gpu. This directory will be generated after the test.
Note* To work with the above-mentioned notebooks change the files available in Changed_Files at appropriate places in UPSNet code.