Arshad Lahounifard
Full-Stack Developer | Senior Computer Vision / AI Engineer
New to Contra
Arshad is building their profile!
✨ Real-time face re-identification on multiple cameras
In this video, I demonstrate the powerful parallel processing of multiple cameras and real-time people identification. My application can handle up to 500,000 faces per second, ensuring maximum efficiency and accuracy. You'll see how the Face Re-Identification feature allows the system to track individuals across different camera feeds, even if they are not facing the camera directly. This advanced capability is crucial for security and surveillance applications. My platform is designed to seamlessly integrate with existing systems, making it easy to implement in various settings. Curious to see it in action? Check out the video now! 👀If you're looking for a cutting-edge face/object recognition solution, let's connect. I'd be happy to discuss how our application can meet your specific needs and take your security to the next level. 🤝 Interested in a customized solution? Let’s connect! 🤝
0
11
Vehicle detection system (count, track, classify, size and speed of vehicle)
Analysis vehicle traffic on highways, intercity roads can be useful to control traffic and make management decisions as well.
Determine the number of vehicles, size, classification and measuring their speed crossing the road is used to better manage the roads and routs. This system is the base on image and video processing technics by helping artificial intelligent tools like deep neural networks.
My Traffic Analyzer software working on both CPU/GPU with capability of change in quality of processing to run on weak hardware system and high level hardware which using GPU.
You can offer your requirement on my software and I will customize it as soon as possible for you.
0
24
Answer sheet checker
Designing answers sheets and surveys and analyzing them is one of the most tedious tasks for researchers and schools. Especially when the number of tests in the sheets is high, the accuracy of the corrector decreases. Oprical mark recognition (https://en.wikipedia.org/wiki/Optical_mark_recognition#:~:text=Optical%20mark%20recognition%20(also%20called,the%20form%20of%20shaded%20areas.)(OMR) software helps correcting many answer sheets automatically and intelligently without user supervision with high accuracy and speed.
Answer sheet designer can easily design different forms and answer sheets then using the answer checker software to analysis photos or scanned from of the answer sheets and automatically the results can be announced as output. In fact, the filled marks for each test are checked by the software with the help of image processing (https://en.wikipedia.org/wiki/Digital_image_processing) and artificial intelligence (https://en.wikipedia.org/wiki/Artificial_intelligence) algorithms.
The ability of software is to define answer sheets with unique barcodes for students which are check by the barcode reader, so you can analysis batch sheets.
0
29
Performance tuning for AI models in production on GPU
• 24 hours of video processed in 16 minutes
AI models are only truly useful when they are fast enough for real-world use.
I recently worked on performance tuning for production AI inference on GPU, with a focus on object detection and high-throughput video processing.
🔥 Demo results:
• YOLO inference at ~1900 FPS on GPU
0
41