AI Shelf Stock Counter POC

₹750-1250 INR / hour

Posted 1 minute ago

•

Ends in 6 days

₹750-1250 INR / hour

I need a quick proof-of-concept that shows I can take a phone photo of a retail shelf, run it through a small vision model, and instantly see how many items are present. The scope is narrow: one image in, one total count out, displayed on a lightweight dashboard. Here is what I’m picturing: • A web page (Python + Flask, FastAPI, or similar) where I upload or drag-and-drop a phone snapshot of a shelf. • The backend calls an image-recognition pipeline (OpenCV, TensorFlow, PyTorch—use what you are most comfortable with) that detects every individual product silhouette and returns the total count. • The response appears on the page as a number; overlay boxes or heat-map visuals are a plus but not mandatory for this stage. • The result, image name, and timestamp should be stored in a simple SQLite or lightweight NoSQL table so I can refresh the page later and see past runs. Acceptance criteria 1. Given a clear, straight-on phone photo of a shelf section, the system returns an item count with at least 90 % accuracy in a test set of 20 images I provide. 2. I can spin the whole thing up on my laptop with one command (ideally Docker compose). 3. Code is clean, documented, and handed over in a private Git repo. If the POC performs well we will move on to identifying product types and spotting empty facings, so keep modularity in mind. Let me know what frameworks and pre-trained models you propose and how fast you can have a demo ready.

Project ID: 40828694

About the project

9 proposals

Open for bidding

Remote project

Active 7 secs ago

Place your bid

Bid amount Email address

Benefits of bidding on Freelancer

Set your budget and timeframe

It's free to sign up and bid on jobs

9 freelancers are bidding on average ₹931 INR/hour for this job

User Avatar
Flag of INDIA

Hi, I can build this POC as a clean, modular image-counting app: a lightweight Flask/FastAPI dashboard for drag-and-drop shelf photos, a Python vision pipeline to detect each product silhouette, and a results view showing the total count with optional boxes/heatmap overlays. I’ll also store image name, timestamp, and count in SQLite for history tracking. For the model, I’d use a proven pre-trained detection approach in PyTorch or TensorFlow, then tune it for straight-on shelf images to improve counting accuracy on your 20-image set. I’ll keep the architecture extensible so we can later add product classification and empty-facing detection without reworking the core pipeline. If this sounds aligned, please start the chat and I’ll walk you through the best framework/model choice and implementation plan. Regards, Sai Bhaskar

User Avatar
Flag of INDIA

Hi I’d build this POC using Python + FastAPI, OpenCV, and a suitable pre-trained object-detection model such as YOLO. The workflow will be simple: upload a shelf image, detect individual products, display the total count, and store the image name, result, and timestamp in SQLite. I’ll keep the detection pipeline modular so product classification and empty-facing detection can be added later. The project will include a lightweight dashboard, optional bounding-box visualization, Docker Compose setup, documentation, and private Git delivery. I have 5 years of full-stack development experience with backend APIs, dashboards, and data-driven applications. I’d estimate 5–7 working days for the initial demo, with the 90% accuracy target validated against your 20 test images rather than assumed upfront. Regards, Ritesh

User Avatar
Flag of VIETNAM

==== Hi - Quy Dao here ==== “ONE IMAGE IN, ONE TOTAL COUNT OUT” - I will build a focused POC around that workflow, with a lightweight upload page, image-counting pipeline, and saved run history. I propose a Python API with a pretrained object-detection model, using OpenCV for image handling and a modular design so product classification and empty-facing detection can be added later. The dashboard will show the count and, if the model supports it reliably, overlay detected boxes. I’ll store the image name, result, and timestamp in SQLite, then package the app for one-command startup with Docker Compose. I’ll document setup and test against your 20 images, comparing counts with the provided references and reporting accuracy. A demo should be ready in about 3–5 days after receiving the images and confirming the run environment; achieving 90% accuracy depends on image quality and how well the pretrained model fits the products. Looking forward to work with you.

User Avatar
Flag of INDIA

I read your project details carefully and can help you with [mention their specific task, e.g., fixing the bug in your Python script / building this software tool] quickly and cleanly. I have strong hands-on experience in Python, SQL, and software development, and I focus on writing well-structured, error-free code. Here is how I will approach your project:

内容来自 Freelancer News,版权归原作者所有。

阅读原文 ↗

用 HomepageTab 新标签页关注 RSS,打开浏览器就能看到更新。

安装 HomepageTab