I will build a custom image classification model using CNN and tensorflow
About this gig
Are you looking for a reliable AI Engineer to build a custom image classification model for your project?
I specialize in building production-ready computer vision models using TensorFlow, Keras, and PyTorch from data preprocessing to a fully trained, evaluated model you can actually use.
What I offer:
Custom CNN model trained on your dataset
Data preprocessing & augmentation pipeline
Full evaluation report (accuracy, F1-score, AUC, confusion matrix)
Clean, documented, ready-to-run code
Optional Grad-CAM explainability visualizations (Premium)
Proven results:
I've built and deployed classifiers including a chest X-ray pneumonia detector (91.67% accuracy, AUC 0.9698) and a satellite wildfire detection system (AUC 0.995, F1 0.96) both live and publicly demoed on Hugging Face.
Why work with me:
- Real deployed projects, not just tutorials
- Clear communication and fast turnaround
- Code + documentation you can actually understand and reuse
Tools I use: TensorFlow, Keras, PyTorch, OpenCV, Scikit-learn
Message me with your dataset details before ordering if you're unsure which package fits your project happy to advise.
Get to know Mansoor Ali
- FromPakistan
- Member sinceJul 2025
- Avg. response time1 hour
Languages
English, Urdu, Hindi
My Portfolio
Other AI Development Services I Offer
FAQ
What format should I send my dataset in?
Images organized in folders by class (e.g., "cat/", "dog/") works best, or a CSV with image paths and labels. I can also work with datasets from Kaggle, Roboflow, or your own collection just share the structure and I'll guide you.
How much data do I need to provide?
For solid results, I recommend at least 200–300 images per class, though I can work with smaller datasets too (results may vary with very limited data). I'll advise on your specific case after reviewing your dataset.
Will I receive the trained model file?
Yes you'll get the trained model file (.h5 or .pt depending on framework), the full source code, and an evaluation report explaining the model's performance.
Can you deploy the model as a web app?
This gig covers model development and evaluation only. If you'd like a full web app with a live interface, check out my other gig for Streamlit/LLM-integrated AI apps.
What if the model's accuracy isn't good enough?
I'll review the results with you and suggest improvements (more data, different architecture, hyperparameter tuning). The Premium package includes an additional implementation round if results need refinement.
Can I see a live demo?
yes https://huggingface.co/spaces/Mansoorrr/chest-xray-pneumonia-detector

