Downloads · 30 days
10
33% of all-time downloads
leixiang25/hw2-tops-sleeve-cnn-automl
hw2-tops-sleeve-cnn-automl is a image classification model from leixiang25. Use it when you need a label for an image. It is set up for keras. The card lists the license as cc-by-4.0.
Classifies an overhead flat-lay photo of a top as shortsleeve or longsleeve. Built for CMU 24-679 Homework 2 to practice budgeted neural-architecture / hyperparameter search. Not a general garment classifier.
Downloads · 30 days
10
33% of all-time downloads
All-time downloads
30
Public
Repo size
5 MB
Likes
0
Public
Click a slice to open those files.
.keras5 MB · 100%
From the Hugging Face model README
Classifies an overhead flat-lay photo of a top as short_sleeve or long_sleeve. Built for CMU 24-679
Homework 2 to practice budgeted neural-architecture / hyperparameter search. Not a general garment classifier.
ypolatog.Rescaling layer), so feed raw 0–255 pixels.{'n_blocks': 4, 'filters': 32, 'dense': 64, 'dropout': 0.07816471529782049, 'lr': 0.00045791512154966046, 'augment': False, 'batch': 16}| Metric | Value |
|---|---|
| Best validation loss | 0.0708 |
| Test accuracy (6 real photos) | 0.667 |
| Majority-class baseline | 0.667 |
Uncertainty: 6 test images → one image ≈ 17 accuracy points. Treat as a rough signal.
Google Colab, NVIDIA T4 GPU; search ≤ 15 min.
Code and this card were drafted with Claude (Anthropic) and reviewed, run, and interpreted by the author.
Dataset by ypolatog (CMU 24-679, Fall 2026).