Book a computer vision consultation
A free 30-minute call with Muhammad Rizwan Munawar to talk through your computer vision or deep learning project. You'll get an honest read on whether it's possible, how accurate it can be, and what it will realistically cost.
Who you’ll be talking to
You'll be talking to me, Muhammad Rizwan Munawar. I'm a computer vision engineer and one of the top contributors to the open-source Ultralytics YOLO project, and I run Rizwan AI.
For the last several years I've built object detection, object tracking, and video analytics systems for startups and teams around the world. The kind that have to keep working outside a notebook, on real cameras and real hardware. I've also written a lot about it, so a fair number of people land here after reading my YOLO and deep learning tutorials.
On the call I won't pitch you. I'll give you a straight answer on whether your idea is doable, how good it can realistically get, and what it takes to ship it.
What I can help with
These cover most of what people ask me for. If your use case isn't listed, send it anyway: a couple of sample images and a sentence about your goal are usually enough for me to tell you whether it's doable and roughly what it would take.
Object detection & tracking
Detect, track, and count objects in images, video, and live camera feeds. I work daily with Ultralytics YOLO (YOLO11, YOLO26), Roboflow Supervision for tracking and zone logic, and trackers like ByteTrack and BoT-SORT.
Training & fine-tuning
Train a model on your own data, or fix one that isn't accurate enough. We cover labeling, dataset cleanup, and getting the model to the accuracy your use case actually needs.
Video analytics
Turn raw camera feeds into numbers you can use: people counting, zone occupancy, dwell time, speed, line crossing, queue length, heatmaps, vehicle and number-plate reading, and more.
OCR & document reading
Pull text, fields, and tables out of invoices, IDs, meters, and forms, including the messy, real-world scans that break off-the-shelf tools.
Edge & production deployment
Get models off the laptop and into production: ONNX, TensorRT, and NVIDIA Jetson, with monitoring and retraining as your data drifts.
A second opinion
Already have a model or a team? I'll review the data, accuracy, and architecture and tell you plainly what to fix, what to keep, and what to drop.
How the call works
Share the context
Tell me what you want to detect, read, or measure, and what data you already have, whether that's images, video, or a live camera feed.
30-minute discovery call
We hop on a quick call and talk it through together: is it doable, what's the likely approach, how accurate it can realistically get, and a fair estimate of the effort and cost involved.
A concrete plan
You walk away with a clear recommendation: a scoped prototype, the right next step, or an honest 'not yet'. Either way, there's no pressure to move forward.
What clients say
These are verified reviews from clients I've worked with on Upwork, across computer vision, deep learning, object detection, instance segmentation, and industrial AI projects.
5 🌟 reviews on- $2,620.80
Excellent to work with and exceeded all expectations in answering questions related to various courses. One of the standout qualities of Muhammed was their commitment to going above and beyond. They didn't just provide brief, surface-level answers. Instead, they took the time to provide in-depth responses that demonstrated their expertise and dedication to the job
VerifiedAugmented AIComputer vision project - $1700 / 55$/hr
Muhammad did an excellent job on my project. He's very knowledgeable, intelligent, and highly capable. I hope to work with Muhammad again in the future.
VerifiedJoe GratzDeep learning project - $600
Even though he had to work till midnight, Rizwan always worked diligently and attempted to solve difficulties. Rizwan is the best freelancer with whom I've ever worked in Upwork.
VerifiedYanuar BramantyoComputer vision project - $317.50 / 35$/hr
Good experience, was available when needed and on time on our appointment. explained well and teached well.
VerifiedDarmandran SivapathasundramComputer vision project - $270
Work was completed promptly and with great care. He's willing to go the extra mile and was willing to adapt to changes. Great job!
VerifiedAnn DelaneyComputer vision project
Hands-On AI: Computer Vision Projects with Ultralytics and OpenCV
Build real computer vision projects from scratch using Ultralytics YOLO and OpenCV.
- Object counting in images and live video
- Queue management and waiting-line analysis
- Track and count objects inside custom zones
- Turn detections into analytical graphs and charts
- Run real-time YOLO inference in a Streamlit web app
From the blog
Tutorials, code, and notes on computer vision, deep learning, and applied AI.
Depth EstimationJuly 27, 20269 min readVideo Depth Anything in Python: consistent depth for video
Depth Anything V2 run frame by frame flickers on video. Video Depth Anything fixes that with temporally consistent depth for long clips. Here is how to run it in Python, with the command line and a reusable script.
YOLO26July 19, 202619 min readYOLO26 vs YOLO11 vs YOLOv8 vs YOLOv10, YOLOv9 & YOLOv5
Six generations of YOLO side by side: published accuracy and parameter counts for YOLOv5, YOLOv8, YOLOv9, YOLOv10, YOLO11 and YOLO26, what actually changed between them, and a straight answer on which one to use for detection, edge deployment, or transfer learning.
OCRJuly 18, 20267 min readHow to extract text from images in Python (OCR): 5 libraries compared
A practical guide to reading text from images in Python: Tesseract, EasyOCR, PaddleOCR, docTR and VLM-based OCR, with runnable code and a which-one-to-pick guide.
Object TrackingJuly 18, 20268 min readHow to use ByteTrack with YOLO for object tracking in Python
Give YOLO a memory. This tutorial uses ByteTrack to assign persistent IDs to objects across video frames, with a runnable Ultralytics script and a bytetrack.yaml tuning guide.
Semantic SearchJuly 6, 20267 min readBuild a semantic image search engine with CLIP and Python
Learn how to build a semantic image search engine that finds pictures by meaning. A few lines of Python turn a folder of images into a searchable index you can query in plain English.
YOLO26July 3, 20269 min readYOLO26 vs YOLO11: Real-time ONNX FPS benchmark in Python
Build one small, reusable class that runs Ultralytics YOLO26 and Ultralytics YOLO11 as ONNX models, draws clean detections, and overlays live FPS and latency so you can compare their real-time speed on the exact same footage.
Depth EstimationJuly 1, 202613 min readDepth Anything V2 in Python: image, video and webcam depth
Learn how to estimate depth from a single image, a video, or your webcam using Depth Anything V2 and a clean, reusable Python class, plus the run.py command, metric depth, and a note on TensorRT.
Object TrackingJune 27, 202613 min readUltralytics object trackers comparison: ByteTrack, BoT-SORT & More
How do the six Ultralytics trackers actually behave on the same footage? A look at BoT-SORT, ByteTrack, OC-SORT, Deep OC-SORT, FastTrack, and TrackTrack, their internals, trade-offs, and side-by-side results on ID switches, ID stability, and FPS.
Frequently asked questions
- Do you work with YOLO and the latest YOLO26?
- Yes. I'm a top contributor to the open-source Ultralytics YOLO project and use it every day. I can help you pick the right version (YOLO11, YOLO26) and model size for your accuracy-vs-speed trade-off, train a custom model on your classes, and export it to ONNX, TensorRT, or CoreML for the device you're deploying to.
- How much does a computer vision project cost?
- It depends on the problem, but I'll give you a real range on the call instead of a vague 'it depends'. A small script or a fine-tune is often a few hundred dollars; a trained model with a clean dataset is usually low thousands; a full production system with deployment and monitoring is more. You'll leave knowing roughly where yours sits.
- How long does a typical project take?
- A proof-of-concept on your own images usually takes a few days to two weeks. Getting from a working prototype to something reliable in production typically takes a few weeks more, mostly spent on data, edge cases, and deployment rather than the model itself.
- I already have a model that isn't accurate enough. Can you help?
- Often, yes. Most accuracy problems come from the data, the labels, or the wrong metric, not the model architecture itself. I'll review your dataset, training setup, and results, then tell you plainly where the real fix is, whether that's relabeling, more data, a different model, or better post-processing.
- Can you deploy models on edge devices like NVIDIA Jetson?
- Yes. I regularly deploy detection and tracking models to edge hardware, including the Jetson family (Orin, Nano) as well as standard GPUs and CPUs, using ONNX and TensorRT to hit the frame rate and memory budget you need. I can also help you decide whether edge or cloud is the cheaper, simpler choice for your case.
- Is my data and project kept confidential?
- Yes. Your images, video, and ideas stay private, and I'm happy to sign an NDA before you share anything sensitive. I won't reuse your data or show your project as a case study without your permission.
- Is the consultation really free?
- Yes. The first 30-minute call is free with no obligation. If it turns out you don't actually need me, or a simple off-the-shelf tool would do the job, I'll tell you that straight.
- What should I have ready for the call?
- A few sample images or a short video clip, a plain-English description of what you want to detect, read, or measure, and any constraints you know about, like camera type, hardware, budget, and deadline. That's usually enough for me to give you a useful answer on the call itself.