From the model card
Fields below are copied from the tags and counters on the HuggingFace repository timm/mobilenetv3_small_100.lamb_in1k at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- timm
- Pipeline tag
- image-classification
- Library
- timm, Transformers
- Framework tags
- PyTorch
- Weight formats
- safetensors
- License tag
apache-2.0— read the license file in the repo before relying on it- Papers cited
- arXiv:2110.00476, arXiv:1905.02244
- Datasets declared
- imagenet-1k
- Downloads (HF counter at last fetch)
- 17,428,712
- Likes (HF counter at last fetch)
- 103
- Model card
- https://huggingface.co/timm/mobilenetv3_small_100.lamb_in1k
Use cases
- On-device image classification for mobile or embedded applications
- Edge vision systems with strict latency and memory budgets
- ImageNet-1k top-level category classification in production pipelines
- Lightweight transfer learning backbone for domain-specific fine-tuning
- High-throughput batch image classification where compute is limited
Pros
- ~2.5M parameters enables mobile deployment and CPU inference
- timm integration provides standardized preprocessing, augmentation, and inference APIs
- LAMB optimizer training improves accuracy at this scale vs. standard SGD
- Apache 2.0 license
Cons
- ImageNet-1k training limits classification to 1000 fixed categories
- Low capacity means accuracy ceiling is substantially below larger models
- Requires fine-tuning for any domain outside natural ImageNet photography
- No bounding box, segmentation, or multi-label output
- timm dependency adds library requirements vs. standalone Transformers models
Tags
timmpytorchsafetensorsimage-classificationtransformersdataset:imagenet-1karxiv:2110.00476arxiv:1905.02244license:apache-2.0region:us