Counting finished products and people involved in the process based on the YOLO model.

Budget: 10000 UAH Deadline: 7 days

Good day! I am doing a count based on video using YOLO — detection and tracking of objects and people, to avoid counting the same instance twice. It is indeed possible to account for products, personnel, and identify defects along your line, plus provide consolidated analytics on downtimes and reconfigurations. One important thing: the video is recorded from cameras above the line, and can you provide a few minutes of recording for accuracy testing? The result depends on the angle and quality of the frame.

Budget: 27000 UAH Deadline: 21 days

This task should not be approached merely as a YOLO model, but as a video analytics system for the production line - with events, accuracy, a stop log, and reports.

The evaluation of the first working stage is from 90,000 UAH and about 21 days. In 10,000 UAH, it seems that only a short technical analysis or hypothesis testing on 1-2 videos can fit, without industrial stability.

I see the implementation software as follows:
> audit of cameras, angles, lighting, and types of products
> preparation of the dataset and training or retraining YOLO for products, people, and defects
> logic for counting without duplications - object tracking, zones, line speed
> stop and reconfiguration events - manual or automatic markers
> analytical dashboard with performance, human participation, defects, and downtimes

Note, there is a nuance here - accuracy will depend not only on the model but also on the quality of the video, camera angle, overlaps, and how clearly the defects are visually visible. A knife is sharp not because it shines, but because it has been sharpened =)

It is necessary to clarify:
> is there already video from the line or access to cameras
> do defects have visual signs in the frame or are additional data needed from the equipment
> is analytics needed as a separate panel or integration into your CRM, ERP, or spreadsheets

Relevant examples:
> https://business.ingello.com/forma-wms - WMS and production-warehouse logic, similar in operation accounting and control events
> https://business.ingello.com/vorfahr - AI automation with data processing and workflow construction around the model
> https://systems-fl.ingello.com/ua - a brief overview of Ingello Systems and our approach to complex systems Expand

Budget: 10000 UAH Deadline: 6 days

Olexiy, for your line, it is important not just to "count objects," but to consistently distinguish between finished products, non-conforming items, people, and moments of stops/reconfigurations. I can design a solution on YOLO with tracking and analytics logic so that you receive accurate throughput and clear reports. I have experience in creating web solutions on Laravel/Vue and in computer science. Let's discuss the video stream, scenarios, and the required accuracy.

Budget: 25000 UAH Deadline: 12 days

We are implementing an intelligent computer vision system based on YOLO architecture (v8/v11) for your conveyor line. I looked at the screenshot from the camera — the angle is perfect for analytics. Since the blue conveyor belt strongly contrasts with the white products, we will be able to ensure the most accurate counting of objects and detection of defects. For tracking without duplication, we will deploy the ByteTrack / BoT-SORT pipeline.

Hello, Oleksiy! I am Nina, the manager of the IT agency Valflow. The development of artificial intelligence systems, computer vision, and video stream processing in Python is handled by Valentyn, a Senior Developer. We create custom CV scripts and clearly understand how to transform an image from the camera into real business analytics.

Our technical plan for your production (according to the frame):
1. Contrast product counting: Thanks to the blue belt and the white color of the blanks, we will set up a line counter. The system will clearly record each unit, and the tracker will ensure that when the belt stops or fluctuates, objects are not counted again.
2. Workforce monitoring: Each worker is located on the line in a defined zone. We will divide the frame into individual ROIs (Regions of Interest) for each workstation. The system will record the presence/absence of a person in the process and the time of their active work.
3. Detection of defects and deformations: We will train the model to classify the geometry or anomalies of the blanks (for example, torn sheets or displacements). For this, we will need to collect a small dataset of defect examples directly from this camera for fine-tuning YOLO.
4. Context of stops: If white objects stop moving on the blue belt, the algorithm will automatically record the line stop, log the timestamp, duration of the downtime, and analyze whether there were people in the reconfiguration zone at that moment, forming a ready analytical report (CSV/JSON/Telegram).

Cost and timelines (for creating a stable working MVP script based on your video recordings):
A full analysis of production lines is an artificial R&D development. The initial stage of analysis and creation of a stable detector:
— Cost: 25,000 UAH (ready to secure in the Safe after detailing the technical specifications)
— Timeline: 10–14 days

Write to me in private messages - we would be happy to take a look at a short video fragment of this line in motion to accurately assess the frame rate (FPS) and discuss the project launch! Expand

Budget: 15000 UAH Deadline: 15 days

Good day!

I have experience in developing AI solutions for automating business processes, data collection, and real-time analytics.

Similar cases:

Development of a monitoring system for production processes, where data from equipment was collected into a single system, analyzing line productivity, downtime, personnel load, and key performance indicators.

AI solutions for quality control, where defects in products were detected using computer vision and analytics on the causes of defects were automatically generated.

For your task, I can implement a system that will collect data from the production line, calculate actual throughput, analyze downtimes and reconfigurations, identify non-conforming products, and generate ready reports for management.

I am ready to discuss data sources (cameras, PLCs, sensors, existing systems) and propose an optimal solution architecture. Expand

Budget: 27000 UAH Deadline: 30 days

I have experience in training very precise models.
The estimated completion time is 1-2 months. Time for creating a prototype, testing, and improvements after testing.

Can you tell me if this is the only camera angle?
And what "defects" are being referred to? Is it really possible to see signs of defects from that distance?

Budget: 10000 UAH Deadline: 7 days

Good day. I can perform YOLO counting of products/people and defects on the line, with tracking and a report on stops. Please send 1-2 short videos from the camera angle — I will let you know what is feasible within the MVP framework. The starting price is negotiable.

Budget: 10000 UAH Deadline: 3 days

Counting based on YOLO for units of products and people in the frame, with classification of defects and recording of stops/reconfigurations along the timeline. To clarify: is the video from one fixed camera, or is support for multiple angles needed? The prototype will be ready in 3 days.

Budget: 10000 UAH Deadline: 14 days

Good day! We have experience in implementing computer vision systems based on YOLO for industrial lines. We carry out product and personnel counting with high accuracy, taking into account the specifics of lighting and conveyor speed. We will configure the model for detecting both quality and non-conforming units of product. We are ready to discuss the details of integration into your infrastructure.

Budget: 10000 UAH Deadline: 14 days

I conducted an analysis of a printing house and a machine engineering enterprise. I will provide you with what is needed as a case. I will invest to the maximum. Write to me.

Budget: 10000 UAH Deadline: 10 days

😀 Hello!
✅ I will perform it with quality 👌
✅ I have done similar work — computer vision in Python (YOLO: detection, tracking, counting) 🎯
🛠 How I will do it: based on YOLO — detection and counting of finished products and people on the line, tracking for precise passage, identification of defects, recording of stops/reconfigurations; analytics/reports (dashboard, CSV). Processing from cameras in real-time or from recordings
🎥 Portfolio: Freelancehunt
✅ Feel free to contact me, I will clarify the line and metrics 👌🙂

Budget: 10000 UAH Deadline: 10 days

Hello! I have experience working with the YOLO model and computer vision tasks. I am ready to implement a product counting system, people detection, and defect classification on your line. To solve the task, I will use Python, OpenCV, and YOLOv8/v9, which will ensure high accuracy and processing speed of the video stream. I will also set up data logging for analytics on downtime and throughput. I am ready to discuss the details of the technical assignment, feel free to write.

Budget: 27000 UAH Deadline: 14 days

We have a practically ready solution for video analytics on YOLO, which can be quickly adapted to your production line and launch the first working phase.

We are in touch and can discuss the details within the marketplace.

For 45,000 UAH and 14 days, I would propose to complete the first phase - a PoC on your videos or on one section of the line. It will include counting finished products, participation of people in the process, basic detection of defects, events of stops or adjustments, and a table with analytics. Full industrial implementation with cameras, server, roles, control panel, and data annotation will likely need to be assessed separately after checking the video stream.

Briefly on the implementation - we take your video samples, check the quality of the frame, areas of interest, line speed, object overlap, and types of defects. After that, we configure the model, counting rules, event log, and output indicators in a convenient form. Look, here’s the nuance - for accuracy, not only the model is important, but also the correct logic of events, otherwise the counter will beautifully make mistakes.

Questions:
- Is there already video from this line for a shift or at least 30-60 minutes of operation?
- Should defects be determined by appearance in the frame or are there separate sensors, markers, or manual confirmation?

Relevant examples:
- https://business.ingello.com/vorfahr - AI and automation of complex logic, close in approach to building an intelligent module
- https://business.ingello.com/forma-bpm - systems for processes, events, roles, and analytics, similar to the part about stops and adjustments
- https://systems-fl.ingello.com/ua - briefly about Ingello Systems and our approach to such systems

To start, we need video fragments, a description of product types, examples of defects, and the desired report format. After that, we will be able to say more accurately where the ready model will suffice and where additional annotation and training are needed. Expand

Budget: 10000 UAH Deadline: 4 days

Good day!! I need more information regarding your project!!!! Thank you!!!

Budget: 10000 UAH Deadline: 7 days

Olexiy, instead of a description — a working prototype, assembled today on stock video of the packaging line:

Google Drive

In the video: product counting through the intersection line (28 pcs in 9 s, ~184 pcs/min), tracking each unit, detecting people in the frame. Simultaneously, the system writes a CSV report: time / quantity / pace / people.

The model — with an open dictionary: product type and non-conformity are set by description, without weeks of labeling. For accuracy on your line, I will train it on footage from your camera — for the pilot, 10–15 minutes of video is enough.

Budget: 10000 UAH Deadline: 14 days

Olexiy, you need to automate the collection of statistics from the production line to eliminate human error in calculations and better understand the reasons for downtime. Using YOLO for real-time object detection will allow not only counting products but also tracking employee activity in the line area.

I will configure the model to recognize your specific objects and the state of the conveyor, after which I will integrate the data output into a user-friendly analytical dashboard. You will receive a system that records transitions between operational states and marks non-conforming items directly during the movement of the belt.

From what angle is the camera usually installed relative to the line, and is there already an accumulated video archive for training the model?