yl0628/autogluon-image-predictor

Model Card: AutoML Neural Network Predictor for Tomato Images

Model Details

• Framework: AutoGluon
• Task: Classification

---

Dataset

• Source: Iris314/Food_tomatoes_dataset
• Target: label
• Splits:
  • Augmented: 490 rows
  • Original: 49 rows
• Preprocessing Steps:
  • Stratify 'label' column.
  • Train/test split (80%/20%).

---

Model

Name	Type	Params	Mode
model	TimmAutoModelForImagePrediction	11.2 M	train
validation_metric	MulticlassAccuracy	0	train
loss_func	CrossEntropyLoss	0	train

Summary

• Trainable params: 11.2 M
• Non-trainable params: 0
• Total params: 11.2 M
• Total estimated model params size: 44.710 MB
• Modules in train mode: 101
• Modules in eval mode: 0
• Validation accuracy: 1
• Training time: ~49.5 seconds

---

Training

• Framework: AutoGluon
• Preset: "medium_quality"
• Image Size: 224x224
• Explored Models: ResNet 18

---

Results

• Test Split:
  • Accuracy: 0.9796
  • Weighted F1: 0.9796

---

Notes

Educational use only. Used AutoML for training model, used ChatGPT and Gemini to debug, used ChatGPT to make table for model info.