ManisAI v1.1

ManisAI v1.1

A Mobilenetv3 AI model that had been developed to classify six types of kuih:

1. Kek Lapis
2. Kuih Lapis
3. Kuih Ubi Kayu
4. Onde-onde
5. Kuih Seri Muka
6. Kuih Talam

This model was trained using a dataset of over 3500 images sourced from the internet, user-submitted photos, and a kuih dataset from Kaggle, and it is trained under Google Colab Notebok's T4 GPU for 2 hours 41 minutes.

Report of the model:

✅ Overall Accuracy: 0.2832

📊 Per-class metrics:

Class	Precision	Recall	F1
Kek Lapis	0.1053	0.0208	0.0348
Kuih Lapis	0.0612	0.0588	0.0600
Kuih Seri Muka	0.0000	0.0000	0.0000
Kuih Talam	0.2569	0.7327	0.3805
Kuih Ubi Kayu	0.3168	0.3232	0.3200
Onde-onde	0.1333	0.0200	0.0348

🎯 ROC AUC per class:

Class	AUC
Kek Lapis	0.5613
Kuih Lapis	0.5283
Kuih Seri Muka	0.3986
Kuih Talam	0.6559
Kuih Ubi Kayu	0.6183
Onde-onde	0.3686

📦 Macro Metrics:

Marco Precision	0.1456
Macro Recall	0.1926
Macro F1	0.1383
Macro AUC	0.4125

Full Changelog