serbekun/CCAiM

CCAiM

overwiev

CCAiM aims to develop an AI-powered model for classifying clouds based on ground-level photographs. The project uses image recognition techniques to identify cloud types according to the WMO International Cloud Atlas classification.

GitHub

The project code for training, inference, and model evaluation is available on GitHub.

Used Datasets

• My dataset on Hugging Face

Models

Model	Accuracy	Macro-F1
V0.0.5 scratch	27.52%	0.123
R18 V0.0.5	42.95%	0.314
V0.0.6 scratch	35.57%	0.237
R18 V0.0.6	50.34%	0.390

Old models

This models was trained when i was not understanding in neurol networks many so this model very bad.

• v_0.0.1 # first model learned by 23 photo.
• v_0.0.2 # second model learned by 42 photo.
• v_0.0.3 # model learned by 88 photo.
• V_0.0.4 # model learned by 165 photo.