| --- |
| license: apache-2.0 |
| base_model: zjs81/Electric-Car-Brand-Classifier |
| tags: |
| - image-classification |
| - transformers |
| - binary-classification |
| metrics: |
| - accuracy |
| - recall |
| - f1 |
| pipeline_tag: image-classification |
| --- |
| |
| # Tesla Binary Image Classifier |
|
|
| A fine-tuned image classification model that answers one question: |
| **Is this car a Tesla? (Yes / No)** |
|
|
| ## Model Description |
|
|
| This model is fine-tuned from |
| [zjs81/Electric-Car-Brand-Classifier](https://huggingface.co/zjs81/Electric-Car-Brand-Classifier) |
| on a custom Tesla vs. Not Tesla image dataset. |
| The classification head was replaced and retrained for binary classification, |
| optimising for **Tesla Recall** to minimise missed detections. |
|
|
| ## Training Details |
|
|
| | Parameter | Value | |
| |------------------|------------------------------------| |
| | Base Model | zjs81/Electric-Car-Brand-Classifier | |
| | Task | Binary Image Classification | |
| | Classes | `tesla`, `not_tesla` | |
| | Learning Rate | 2e-5 | |
| | Batch Size | 8 | |
| | Epochs | 5 (best at Epoch 4) | |
| | Optimizer Metric | Tesla Recall | |
| | Inference Threshold | 0.6 | |
|
|
| ## Evaluation Results (Test Set) |
|
|
| | Metric | Score | |
| |---------------------|--------| |
| | Accuracy | 95.0% | |
| | Precision (Tesla) | 93.3% | |
| | Recall (Tesla) | 97.0% | |
| | F1 Score (Tesla) | 95.1% | |
| | F1 Score (Not Tesla)| 94.9% | |
