Cattle Weight Estimation from Phone Photos for Small Dairy Farmers and Livestock Buyers

The problem: guessing weight

Most dairy farmers in Pakistan keep a few animals. Weight matters to them every week: for feed planning, for deworming and medicine doses that depend on body weight, for checking that a calf is growing, and for selling at a fair price. Very few own a livestock scale. They guess, or use a measuring tape around the chest and a formula, which takes two people and a calm animal.

At livestock markets, especially before Eid ul Adha, buyers and sellers estimate weight by eye, and prices follow those guesses. A quick, honest estimate from a phone would help both farmers and buyers. It would also give a farmer a record over months, which no amount of guessing provides, and make it easier to notice an animal that has stopped gaining weight before it falls ill.

What already exists and the gap

• Livestock scales and weighbridges are accurate but expensive and found on large farms and some markets.
• Heart girth tapes with a formula are widely used by farmers and vets as a low-cost estimate.
• Research systems estimate cattle weight from photos, depth cameras or 3D point clouds, including several smartphone-based studies published since 2024.

The gap is an app designed for local breeds and conditions, such as Sahiwal and crossbred cows and buffalo in crowded yards, that shows a range instead of a falsely precise number, keeps a growth history and is validated against a real scale.

Inspiration

"Cattle weight estimation model through readily photos" by L. Bai, C. Guo and J. Song, published in Engineering Applications of Artificial Intelligence in 2025, proposed a lightweight network deployed on mobile phones that finds keypoints in cattle photos, derives body measurements and estimates weight. Your project follows the same keypoints-then-measurements approach, which is easier to check and explain than predicting weight directly from pixels.

What you build

• Farmers photograph an animal beside a printed marker and get a weight range, saved to that animal's record and growth chart.
• Vets and livestock officers use the range to sanity-check doses and track herd growth.
• Buyers and sellers get a neutral estimate to start a price discussion.

The result screen shows the photo with the detected keypoints and measurements drawn on it, so the user can see what the app measured and spot an obvious mistake, such as a point placed on a rope or a fence post. Each animal gets a simple record with a name or tag number, and the growth chart compares monthly estimates. A farmer who sees a calf's line flatten knows to check its feed or health early.

The technically interesting part

• Scale from a single photo. A photo has no scale. A printed marker of known size placed beside the animal, at the same distance from the camera, gives one. Measure how errors grow when the marker is closer or further than the animal.
• Pose matters. A cow turned slightly towards the camera looks shorter. Detect pose from keypoints and reject photos that are not side-on.
• Keypoints to weight. Fit a simple regression from body length, height and estimated chest depth to weight on your scale data, and compare with the heart girth formula on the same animals.
• Ranges, not numbers. Use your validation errors to show a likely range, and make the range wider when the photo is poor.

How to evaluate it

• Weight error: mean absolute error and percentage error against the scale, by breed and size.
• Comparison: the same measure for the heart girth tape method on the same animals.
• Range honesty: share of animals whose true weight fell inside the shown range.
• Field usability: time per animal and share of photos accepted first time.

Suggested plan

• Weeks 1 to 3: read the paper, partner with a veterinary university, government livestock farm or large dairy farm with a scale.
• Weeks 4 to 7: photograph and weigh animals, label keypoints, train the keypoint model.
• Weeks 8 to 10: measurements, regression, ranges and the capture screen.
• Weeks 11 to 13: app records and growth chart, field test.
• Weeks 14 to 15: evaluation, report and demo.

Path to a product

Dairy companies that buy milk from small farmers, livestock insurers, microfinance lenders who lend against animals and market operators all need weight estimates. A free farmer app with paid tools for these businesses is realistic. The project shows applied computer vision with careful validation in a sector few developers serve.

Scope warning

Do not promise scale-level accuracy; show ranges and say how they were measured. Never photograph animals without the owner's permission. Keep to adult cattle and buffalo in the main scope. Medicine dosing must always point to a vet's advice, never replace it.

Learn what you need on Sefism

• Mobile development roadmap: camera, offline records and charts.
• Machine learning roadmap: keypoint models, regression and evaluation.
• Ultralytics YOLO Documentation: training a pose model for animal keypoints.
• Label Studio: labelling keypoints on your photos.
• Flutter Documentation: building the app.

Get help with this idea

The risk is a model trained on too few weighed animals. Tauseef can help you find a partner with a scale, plan data collection and present the error honestly. If you are new here, create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session