MMPretrain

Google Colaboratory - Image Classification with Pretrained ResNet18 on Version 1.2.0
https://colab.research.google.com/drive/1igoW8ryWYXDkcAExicaciFvHHFbscHDa

Google Colaboratory - Custom Dataset Training with ResNet18 on Version 1.2.0
https://colab.research.google.com/drive/1RkrPGyAcS0c6CH3eC07EKKwYrzm7ovlX

ml_dataset_binimage_mmpretrain.tar.gz

See MMDetection also for tricks and traps

https://github.com/open-mmlab/mmyolo/issues/1013

python mmpretrain/tools/visualization/vis_cam.py ...
## TypeError: forward() got an unexpected keyword argument 'use_cuda'

The latest version of grad-cam 1.5.2 is not compatible by missing the 'use_cuda' argument (Confirmed 2024/07/11)
Installing version 1.3.6 as follows fixes this problem

pip install 'grad-cam==1.3.6'

https://github.com/open-mmlab/mmpretrain/issues/101

python mmpretrain/tools/train.py ...
## KeyError: 'img'

dict(type='LoadImageFromFile') is necessary in pipelines if loading image data from not pickle but image files

https://mmpretrain.readthedocs.io/en/stable/_modules/mmpretrain/apis/model.html#inference_model

model = get_model(config_file, pretrained=checkpoint_file, device=device)
result = inference_model(model, img)
## No available inferencer for the model

metainfo.results.result.tasks needs to be defined in the model to use inference_model()
The following alternative approach is available for e.g., Image Classification

from mmpretrain.apis import ImageClassificationInferencer
inferencer = ImageClassificationInferencer(config_file, pretrained=checkpoint_file, device=device)
result = inferencer(img)

https://mmpretrain.readthedocs.io/en/latest/user_guides/dataset_prepare.html
https://mmpretrain.readthedocs.io/en/latest/user_guides/train.html
https://mmpretrain.readthedocs.io/en/latest/notes/pretrain_custom_dataset.html
https://mmpretrain.readthedocs.io/en/latest/notes/finetune_custom_dataset.html

https://github.com/open-mmlab/mmpretrain/blob/main/mmpretrain/models/losses/__init__.py

https://pytorch.org/docs/stable/optim.html

optimizer=dict(_delete_=True) removes originally-defined parameters in dict()

https://github.com/open-mmlab/mmpretrain/blob/main/mmpretrain/models/backbones/resnet.py

• frozen_stages = -1 : not freezing any parameters
• frozen_stages = 0 : freezing the stem part only
• frozen_stages = n [n >= 1] : freezing the stem part and first n stages

https://github.com/open-mmlab/mmpretrain
https://mmpretrain.readthedocs.io/
https://github.com/open-mmlab/mmpretrain/issues/1474
https://www.kkaneko.jp/ai/win/mmclassification.html
https://www.kaggle.com/code/haoqingsong/brain-tumo-classification-swinv2-accuracy-100

Daiphys is a professional-service company for research and development of leading-edge technologies in science and engineering.
Get started accelerating your business through our deep expertise in R&D with AI, quantum computing, and space development; please get in touch with Daiphys today!

Daiphys Technologies LLC - https://www.daiphys.com/