Enhance PoolShark ball Detection and accuracy

@MuhammadUzair16

Hi I am a software engineer with over 16 years of experience. I have extensive experience with C++, Qt, OpenCV, image-processing pipelines, and practical computer-vision systems affected by changing lighting and camera conditions. I can review PoolShark's current detection and colour-classification pipeline, reproduce the misclassification cases, and improve it using appropriate masking, illumination normalization, calibrated colour spaces, and confidence-based classification. I will also check tracking and temporal consistency so recognition remains stable throughout a complete game rather than working only on isolated frames. Where useful, I can add diagnostic views and tunable parameters to make calibration and future testing easier. Do you have representative game videos or images showing the failures under different table lighting, and is the camera position fixed? Please contact me to discuss details.

@intelliresponse

Hi, I can help improve the PoolShark ball colour recognition and make the detection more reliable throughout a full game. I’ll review the existing Qt 5.12/OpenCV implementation, identify why colours are being misclassified, and improve the masking, lighting correction, colour-space handling, and classification logic. I’ll also test the changes against different lighting, shadows, and ball positions to make sure the results are more consistent. I’m comfortable working directly with the existing GitHub code and can start by reproducing the current issues. Thanks!

@anandinfraveo

**Free to dig into PoolShark right now, I run Qt 5.12 plus OpenCV builds daily** I am Anand, full-stack dev with infra and automation work for small teams that need someone to take it and run. I can tighten your colour classifier with better masking plus illumination correction and move it to HSV/LAB with calibrated thresholds so reds and oranges stop swapping over a full game. Background: 4.9 star rating, dozens of reviews, projects from server migrations to full apps built from scratch. Recent work: pool table tracker with OpenCV colour calibration, Qt camera app with live object detection overlay. Can you share a short game video with your current lighting so I can test masking fixes right away?

@sf264

Hi there, I understand you need to improve PoolShark’s existing ball detection system and increase accuracy using OpenCV. I can help refine the detection pipeline, reduce false positives and missed balls, and improve position tracking through image processing, contour filtering, and temporal tracking. I’ll focus on improving your existing implementation and validating the results against your actual footage. Could you share the current code and a sample video so I can identify the main accuracy issues? Best regards, Fahad.

@emilesd617

Hi, I am a C++ computer vision developer with 8 years of rich experience in software development, with a background in real-time object detection and colour classification. I am familiar with C++, Qt 5, OpenCV, image processing, colour spaces (HSV/Lab) and object tracking, etc. To improve accuracy, I'd first normalize the lighting by using the table felt as a white-balance reference and applying CLAHE. Then I'd mask each detected ball to its inner area, excluding specular highlights and edge shadows, which usually cause most colour mislabels. I'd classify in Lab space using a hue histogram plus the ratio of white pixels, which also separates solids from stripes. A quick calibration at the start of a game would capture reference colours for your table and lighting. Finally, tracking each ball across frames would smooth out single-frame errors, so labels stay stable through a whole game. I'm an individual freelancer and can work on any time zone you want. Please contact me with the best time for you to have a quick chat. Looking forward to discussing more details. Thanks. Emile.

@bohdan38

Hi, Improving PoolShark for an entire game requires more than tweaking thresholds; colour decisions must remain stable across lighting changes, table backgrounds, shadows, and partially occluded balls. The implementation would focus on stronger ball masking, illumination normalization, and testing HSV or Lab-based classification against representative frames. I would also inspect confidence handling and misclassification patterns so ambiguous detections are rejected or refined rather than propagated. Similar work includes computer-vision enhancements involving image preprocessing, segmentation, and reliable object classification. I can work directly within the existing Qt 5.12 and OpenCV codebase while preserving its current workflow. Within two days, I can deliver a focused working improvement, document the changes, and validate results across varied game frames. The first milestone will target measurable colour accuracy gains and a clean path for further tuning. Best regards, Bohdan

@emanz03

Colour errors in pool footage usually come from lighting shifts, highlights and averaging mixed ball/background pixels rather than detection alone. I’d benchmark the current classifier on representative frames, tighten the ball mask, normalize illumination, and compare HSV/Lab plus robust colour statistics instead of relying on one raw average. I’d also validate tracking across full-game footage so labels stay stable as balls move through different lighting.

@elijahhmonreal

Hi there, I see you're working on improving the ball detection accuracy in PoolShark, a challenging computer vision problem that requires precision colour classification and robust handling of real-world lighting conditions. I'm a Computer Vision Engineer with 12+ years of experience building production-grade image processing systems. I've spent extensive time optimizing colour classifiers, implementing advanced masking techniques, and deploying illumination-invariant algorithms across various applications. For your poolShark enhancement, I'll focus on three key areas: refining the colour-space strategy (exploring HSV, LAB, or adaptive models), implementing adaptive masking to isolate balls more reliably, and applying illumination correction to handle inconsistent lighting throughout a full game. I'll systematically test each enhancement against your complete game footage to ensure accuracy across real-world conditions. Let me know if my profile looks interesting, and we can set up a time to talk. Best regards, Elijah M

@smitk9999

I understand you want PoolShark's ball recognition to stop mixing up colours, so the glasses can follow every ball correctly through a whole game of pool, using better masking, lighting correction or a different colour approach in the existing Qt 5.12 and OpenCV code. I've worked with OpenCV in C++ on colour segmentation and object detection, and since the camera is on glasses that move around the table, I'd rebalance each frame against the table cloth's colour before classifying, then tell stripes from solids by how much white each ball shows. I can start work right away and contact me to discuss more details. Thanks. Smit

@markom435

Hi there, The PoolShark ball detector is mislabeling colors too often, especially across changing lighting and a full pool game. I have strong expertise in solving exactly this type of computer-vision problem through C++ Programming. I will inspect the current OpenCV pipeline, improve masking and illumination normalization, compare HSV/Lab and other color-space strategies, tune classifier thresholds using representative frames, and validate tracking accuracy across varied table conditions in Qt 5.12. Best regards, Marko

@brenob18

Hi. I understand you want the ball recognition in PoolShark to stop mislabeling colors so often, with a sharper color classifier and whatever other improvements it takes to track balls reliably through a whole game, keeping it on your current Qt 5.12 and OpenCV setup. I am familiar with C++, Qt, and OpenCV, and have experience developing real-time object detection, color segmentation, and tracking for moving camera footage. Based on my previous experience, I think most of the mislabeling comes from lighting shifts as the glasses move around the table, plus close color pairs like red vs. orange vs. maroon and yellow vs. the cue ball, so I'd add illumination correction using the felt as a reference, switch classification to Lab or HSV with per-ball calibration, and sample only the inner region of each ball to cut glare and edge noise. I'd also handle stripes vs. solids by measuring the white-to-color ratio, and add frame-to-frame tracking so a ball's label stays stable instead of flickering. I am ready to start the project right away. Best Regards. Breno.

@ayesha86664

Hi there! I understand the main issue is unreliable pool ball colour recognition, and you need the classifier to stay accurate throughout a full game under changing lighting and table conditions. I have experience with C++, Qt, OpenCV, image processing, segmentation, masking, colour classification, and computer vision optimization. I can work directly with the existing Qt 5.12 and OpenCV codebase. I will review the current detection pipeline and identify where colour classification is failing. I can improve masking, apply illumination correction, test HSV/Lab or other suitable colour spaces, and tune the classifier using real frames from your application. I will also test different ball positions and lighting conditions to improve consistency across a complete game without breaking the existing detection flow. check our work https://www.freelancer.com/u/ayesha86664 Could you share a few sample frames where the current classifier mislabels the ball colours? Let me know if you’re interested & we can discuss it. Best Regards Ayesha

@amadorjay

Hi there, The hard part is not just implementing the colour classifier; it’s achieving consistent accuracy across various lighting conditions and ball placements. Improving the masking techniques and adjusting for illumination changes will be crucial to enhance the classifier’s reliability. Additionally, exploring alternative colour spaces could yield better results for distinguishing the balls. To address the specific issues, would you prefer focusing primarily on refining the current classifier, or are you open to incorporating new methods for overall enhancement? Thank you.

@tobiask5

Hi, This is primarily a computer-vision classification problem: PoolShark already runs with Qt 5.12/OpenCV, but the detection pipeline is not robust enough under real pool-table lighting and changing ball appearances. I’ll inspect the current segmentation and classifier, improve masking and illumination handling, test suitable colour-space/features, then validate accuracy across different frames and game conditions. The main risk is that fixed colour thresholds may work in controlled images but fail with shadows, reflections, table colour, and changing camera exposure. Can you provide a few representative game recordings or image samples where the current classifier mislabels the balls? Thanks, Tobias

@richardb063

Hi, starting with the colour classifier in `classifyHsv` and the ball masking pipeline that feeds it. Getting your ball colours labelled correctly across a full game is the goal. Previously I have fixed OpenCV colour classifiers where fixed HSV thresholds broke under changing table lighting, which is exactly what happens with recording glasses as your head moves. I would start with illumination correction, since your current thresholds assume stable light and that is the main source of drift. Then tighten the ball mask so highlights, shadows, and the number stripe do not drag the average colour, and consider Lab or YCbCr where the hue ranges overlap, particularly red against orange and blue against purple. Testing would use your own captured frames, since felt and ball colours are what matter here, not a generic dataset. I am determined to win this project and confident I can deliver high quality results within your deadline if awarded. Talk soon.

@arimua

A pool vision system is only useful when it can reliably tell one ball from another. Hello, I’m an experienced C++ developer with OpenCV and Qt experience, and I can improve the existing poolShark project rather than rebuilding it unnecessarily. I’ll focus on stronger masking, illumination handling, colour space tuning, and classifier accuracy across a full game. I’ll test against varied lighting and table conditions, measure misclassifications, and keep the solution stable with Qt 5.12. I can also improve detection robustness where useful. Let’s make the ball recognition reliable enough for real games.

@LeonBoshoff

Enhancing the PoolShark ball detection sounds like an exciting challenge. I have recently worked on projects involving OpenCV and Qt, focusing on improving object recognition and color classification. To address the mislabeling of colors, I can implement advanced techniques such as improved masking and illumination correction, alongside exploring alternative color-space strategies that can significantly enhance accuracy throughout an entire game of pool. I understand the importance of precision in your application, and my practical approach ensures that the solutions I provide align with your project goals. I prioritize reliability and clear communication, making the development process smoother. What specific lighting conditions are you encountering that affect detection? If you don’t choose to work with me, you’ll still benefit from a free consultation with an expert in the field. P.S I look forward to your message Leon B.

@ankits0192

Hi there, I understand you already have PoolShark running with Qt 5.12 and OpenCV, but the current colour classifier is still mislabeling balls too frequently and needs to become reliable throughout a complete game. The main challenge is handling real-world variations such as table colour, shadows, reflections, camera noise, and changing illumination rather than relying only on fixed RGB thresholds. I’d first reproduce the current detection pipeline, identify where classification errors occur, then improve masking and segmentation and evaluate HSV, Lab, or another suitable colour space with calibrated thresholds. I’d preserve the existing Qt/OpenCV architecture where practical, test against representative frames, measure classification accuracy, and make targeted changes without disrupting the existing ball detection and positioning logic. The final code would remain clean and maintainable. Q1 – Can you provide a few frames or short video showing the current colour misclassifications? Q2 – Is the camera, table and lighting setup fixed during normal use? Q3 – Do you have expected colour labels for the balls so improvements can be measured objectively? Looking forward to making PoolShark’s ball recognition significantly more reliable. Thanks

@millz

Small note first: it'll help me a lot to see a few sample frames from your actual table, because lighting is usually the whole problem here. The core issue with pool-ball colour classification is almost always the colour space, not the classifier. If it's working in RGB, then a shadow across the table or a bright spot from an overhead lamp shifts all three channels at once, so the same ball reads as two different colours depending where it sits. Converting to HSV and classifying mainly on hue, which barely moves under changing brightness, fixes the bulk of the mislabels on its own. The stripes need handling as their own case too, since a striped ball is mostly white and a naive average colour lands somewhere between white and the stripe. The one that HSV alone won't solve is the pairs that share a hue, the reds and the maroon 7, yellow and the orange 5. Those usually need the classifier calibrated to your specific table and lighting once, sampling the real ball colours rather than assuming ideal values, because "yellow" under a warm bulb isn't the textbook yellow. Background on me: 15+ years across C/C++ and Python, including OpenGL and Qt graphics work on CAD and drilling-visualisation software, and I run a self-hosted vision pipeline on OpenCV, so the masking and colour-space side is familiar ground. I'll be straight though, I haven't shipped this exact repo, so my first step is building it as-is and reproducing the mislabels before changing anything.