We're relaunching with a new vision in May 2026! Learn More
Shipped

Rent2Reuse Demo Model

Fine-Tuned MobileNetV2 Image Classification Model for Prohibited and Accepted Items: This project showcases a fine-tuned MobileNetV2 image classification model trained to distinguish between Prohibited and Accepted items. The dataset includes 180 distinct classes, with 35 categorized as prohibited and 145 as accepted. The model was trained using TensorFlow on a GPU-accelerated environment with WSL. To enhance usability, I integrated Gradio for an interactive web interface, providing both top-1 and top-3 predictions with corresponding confidence scores.
Rent2Reuse Demo Model
imzhyke
Ezekiel P. Villadolid ·
March 11, 2025
·
TensorFlowMobileNetV2GradioHugging Face SpacesPythonWSLJupyter Lab
· Live Demo ↗

The Journey

💡
Learning
🔹 Training & Fine-Tuning: The model is built using MobileNetV2 and trained to classify items into Prohibited and Accepted categories. It handles 180 classes, with 35 prohibited and 145 accepted. 🔹 Interactive Demo: I integrated the model with Gradio to create a simple web interface, now live on Hugging Face Spaces. 🔹 Key Learnings: ✅ Set up TensorFlow GPU on WSL for faster training. ✅ Used Git LFS to manage large model files during deployment. ✅ Added top-3 predictions to improve result insights.
📝

Lessons Learned

Learned fine-tuning techniques using MobileNetV2 to achieve improved classification accuracy. Gained experience in GPU setup and troubleshooting on WSL for efficient TensorFlow model training. Learned to build and deploy interactive model interfaces using Gradio for enhanced user experience. Developed strategies to categorize predictions into distinct groups for better interpretation of results. Explored best practices in managing large model files and ensuring proper deployment workflows on Hugging Face Spaces.

What do you think?

Leave a Comment

Comments