STELLAR-CLASS-5L-8-E1 (Custom CNN, no final pooling) - Fetal Planes Classification

This model is a custom 5-block convolutional network, trained from scratch (no ImageNet pretraining) on single-channel 128x128 fetal ultrasound images, to classify fetal ultrasound standard planes.

Model Performance (Test Set)

• Accuracy: 0.2733
• Macro F1 Score: 0.2364
• Macro Precision: 0.4507
• Macro Recall: 0.2889

---

Dataset Information

This model was trained on dataset Marc-HealthAI/combined_fetal_ultrasound_dataset.

• Classes: Fetal abdomen, Fetal brain, Fetal femur, Fetal spine, Fetal thorax, Maternal Cervix, Other, Placenta
• Dataset Features: ['image', 'label']

Dataset Description

No explicit description provided in dataset metadata. Changes Made Removed the pooling on the final layer. Lowered lr = 1e-5 effective batch size 32

Original Dataset README Content

Click to expand original dataset README