| # Model Card for scottymcgee/image-classifier-stop-sign |
|
|
| This model classifies traffic-sign images as either **containing a stop sign** or **not containing a stop sign**. |
| It was trained with AutoGluon’s `MultiModalPredictor` on a binary image dataset of street signs. |
|
|
| ## Model Details |
|
|
| ### Model Description |
|
|
| - **Developed by:** Scotty McGee |
| - **Model type:** Image classifier (binary classification) |
| - **Languages (NLP):** Not applicable (vision model) |
| - **Finetuned from model:** Timm image backbone used by AutoGluon (default is EfficientNet or ResNet depending on config) |
|
|
| ### Model Sources |
|
|
| - **Repository:** https://huggingface.co/scottymcgee/image-classifier |
|
|
| ## Uses |
|
|
| ### Direct Use |
|
|
| Use this model to classify whether an input image contains a stop sign or not. It takes an RGB image as input and returns a predicted label and probabilities. |
|
|
| ### Downstream Use |
|
|
| It can be incorporated into larger perception systems (e.g., driver assistance, robotics) as a pre-screening classifier. |
|
|
| ### Out-of-Scope Use |
|
|
| Not intended for: |
| - Safety-critical deployment without further validation. |
| - Identifying other sign types beyond stop / no-stop. |
| - High-stakes enforcement or surveillance applications. |
|
|
| ## Bias, Risks, and Limitations |
|
|
| The model is trained on the specific dataset you provided. It may: |
| - Misclassify unusual or occluded stop signs. |
| - Perform poorly on non-U.S. stop sign shapes/colors if not present in training. |
| - Inherit any biases in the training images. |
|
|
| ### Recommendations |
|
|
| Always test on your target data before deployment. Combine with additional checks in safety-critical scenarios. |
|
|
| ## How to Get Started with the Model |
|
|
| ```python |
| from autogluon.multimodal import MultiModalPredictor |
| |
| predictor = MultiModalPredictor.load("scottymcgee/image-classifier-stop-sign") |
| preds = predictor.predict(["example.jpg"]) |
| print(preds) |
| |
