Browse categories
Explore
Fiverr Pro
English
$
USD
I will develop a custom model for detection classification and segmentation
M Ihtesham Khan
About this gig
You have a computer vision problem that needs a real solution not copied code from tutorials. I build high-performance AI systems for object detection, tracking, classification, and image segmentation that work in real-world environments.
I am a PEC Registered Software Engineer (COMP/028521) with hands-on experience in YOLO, OpenCV, PyTorch, and TensorFlow. I developed a real-time traffic violation detection system with helmet detection, seatbelt monitoring, tinted window analysis, and Automatic Number Plate Recognition (ANPR) running on live multi-stream video.
WHAT I CAN BUILD
Custom CNN / YOLO object detection
Image classification with CNN, ResNet, EfficientNet
Real-time object tracking using ByteTrack & OC-SORT
Image segmentation with U-Net & Mask-RCNN
ANPR / License Plate Recognition systems
OCR pipelines for documents and ID cards
Face detection and recognition systems
Video analytics and surveillance AI
TensorFlow Lite deployment for mobile/edge devices
FastAPI APIs for real-time inference
Docker and cloud deployment (AWS, Hugging Face)
Streamlit dashboards for visualization
TECH STACK
Python | YOLO | OpenCV | PyTorch | TensorFlow | Keras
FastAPI | Docker | AWS |
Get to know M Ihtesham Khan
M Ihtesham Khan
AI, ML, DL, Computer vison, NLP, Transformers, Chatbots,
- FromPakistan
- Member sinceMar 2021
- Avg. response time1 hour
Languages
Pashto, Urdu, Hindi, English
Hi! I'm Engr. Muhammad Ihtesham Khan, a professional Mobile App Developer with expertise in building scalable, high-performance iOS and Android applications. I specialize in Flutter, React Native, and Native development — delivering apps that are fast, beautiful, and fully functional. From UI/UX design to App Store deployment, I handle the full cycle so you don't have to worry about a thing. Let's bring your app idea to life!
FAQ
Do you work with custom datasets or only public ones?
I work exclusively with your custom dataset. Whether you have 200 images or 20,000, I handle the full pipeline annotation guidance, preprocessing, augmentation, training, and evaluation. If your dataset needs cleaning or labelling, we can discuss that as an add-on.
What format will the final deliverable be in?
You receive clean Python source code, the trained model weights (.pt / .h5 / ONNX), an inference script, a README with setup instructions, and a short video demo of the model working on test data. Everything is packaged so you can run it immediately.
Can you deploy the model so it runs on a website or mobile app?
Yes. I can wrap the model in a FastAPI REST endpoint, containerise it with Docker, and deploy it to AWS or Hugging Face Spaces. For mobile, I convert models to TensorFlow Lite for on-device Android inference I have done this in my own published Play Store apps.
My dataset is very small. Can you still get good results?
Small datasets are a speciality of mine. I use transfer learning from ImageNet-pretrained models, aggressive augmentation strategies, and class-balancing techniques that dramatically improve performance on limited data.I will be honest with you if the dataset is genuinely too small to produce result
Do you provide support after delivery?
Yes — I offer 7 days of post-delivery support for setup issues, bug fixes, and clarification questions at no extra cost. Extended support packages are available as a gig extra.
