Thai snake image classifier

The trained checkpoints of the Thai snake identification benchmark are in benchmark_v4/: eleven final models for 139 Thai snake species (nine architectures at 224 px and ConvNeXt-V2-Base at 384 px with cross-entropy and with Seesaw loss), each with its configuration and locked-test scores. Records, predictions and code are in https://github.com/kritsong/thai_snake_classification (folder benchmark/), and an interactive demo is at https://huggingface.co/spaces/kritaphatson/thai-snake-image-classifier-app.

The ConvNeXt-V2 checkpoints derive from weights released under CC-BY-NC-4.0 and are for non-commercial use. Frozen image embeddings of all benchmark images, for training and scoring classifiers on the exact split, are in benchmark_v4/embeddings/.

The files at the root of this repository are an earlier model, kept for reference.

Downloads last month
19
Safetensors
Model size
86.8M params
Tensor type
I64
Β·
BF16
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Spaces using kritaphatson/thai_snake_image_classifier 2