|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
- bertscore |
|
|
library_name: keras |
|
|
tags: |
|
|
- dogs |
|
|
- cats |
|
|
- mozie |
|
|
- detection |
|
|
pipeline_tag: image-classification |
|
|
--- |
|
|
# 🐾 MozieFinder |
|
|
|
|
|
**MozieFinder** is a lightweight convolutional neural network (CNN) model built from scratch using **TensorFlow**, designed to classify images as either **cats or dogs**. |
|
|
|
|
|
- 📦 **Model Type**: Custom CNN (ResNet-inspired) |
|
|
- 🐶🐱 **Task**: Binary Image Classification (Cat vs. Dog) |
|
|
- 🧠 **Trainable Parameters**: ~1.2 million |
|
|
- 🖼️ **Input Resolution**: 224x224 |
|
|
- 🏋️ **Training Data**: ~20,000 labeled cat and dog images |
|
|
- 🎯 **Validation Accuracy**: ~92% |
|
|
|
|
|
MozieFinder was trained from scratch — no pre-trained weights were used — as a demonstration of how to build a robust image classification model end-to-end. |
|
|
|
|
|
> ⚠️ **Disclaimer**: This model card was written by the model creator. It has not been officially reviewed by TensorFlow or affiliated teams. |
|
|
|
