Can you count calories from a photo?

Roughly, yes. A photo app can usually name what's on your plate and give you a ballpark calorie number, but that number is often off, mainly because judging portion size from a flat picture is hard and because cooking oil, sauces and drinks are easy to miss. Treat a photo log as a quick estimate: check the portions, add what the camera can't see, and judge your progress by weekly averages rather than any single meal.

Below is how photo logging works, what the research says about its accuracy, what it tends to miss, and how to get closer to the real number without weighing every bite.

How photo calorie counting works

Whatever the app, a photo goes through three steps, and each one can add error:

1. Identifying the food. The software decides that the beige pile is rice, not couscous, and that the drink is cola, not iced tea.
2. Estimating how much there is. It guesses the size of each portion from the picture.
3. Looking up the nutrition. It matches each food and amount to an entry in a food database.

The second step is the hard one. A photo is flat, so the app has to infer the height of a pile of pasta or the depth of a bowl of soup. A 2023 systematic review in Annals of Medicine found that estimating volume and calories was the least-researched part of the process, and the authors noted problems such as the lack of depth information in a normal photo. Research tools often put a reference object, such as a marker of known size, next to the plate for exactly this reason.

How accurate is it? What the research shows

A recent test of four apps found calories about a third too low

The most direct test so far came from researchers at the National Institute of Diabetes and Digestive and Kidney Diseases, presented at the American Society for Nutrition's NUTRITION 2026 meeting in July. They took standardized photos of 102 meals made in a research kitchen, where every ingredient is weighed to the nearest 0.1 gram, and asked four popular photo-based calorie apps to estimate them.

Across all four apps, the estimates came in about 250 to 345 calories too low per meal on average, and fat was underestimated by about 30 grams. Carbohydrate estimates were more consistent than the rest. In a follow-up set of more than 200 meals, high-fat ketogenic meals were the hardest to get right.

Keep the limits in mind. This was a conference abstract that hadn't yet gone through peer review, the meals came from a clinical diet study that included high-fat meals most people don't eat every day, and apps update their software often. The researchers' practical advice still holds: if you use a photo app without adjusting the portions or entering amounts, expect your real intake to be higher than the app says, especially for fat.

What reviews of many studies say

The 2023 Annals of Medicine review looked at 52 studies that compared fully automated photo analysis with a reference, such as weighed food. Average calorie errors ranged from 0.10% to 38.3%, which shows how much results depend on the tool and the test. Errors were smaller for photos of single, simple foods than for plates of several foods, and the authors flagged mixed dishes like curries, and plates of overlapping foods, as higher risk for errors. The studies were too different to combine, and the authors concluded that the tools need more development before they can be used as stand-alone dietary assessment.

A 2020 review in the International Journal of Obesity covered a broader range of photo methods, including ones where trained staff, not software, analyze the pictures. Its conclusions are useful for anyone logging at home:

• Trained human analysts were more accurate than automated analysis. In one method, the Remote Food Photography Method, 50 adults photographed their food for six days. Their estimated intake was on average only 3.7% below their intake as measured with doubly labeled water, a gold-standard lab method. The day-to-day spread was wide, though, with a standard deviation of about 29%. (The review's authors hold intellectual property for that method.)
• Fully automated analysis wasn't there yet. The authors wrote that fully automated assessment "with acceptable precision" was "not yet a reality."
• Missing photos are a big source of error. In one study of pregnant women, photos captured only about 64% of measured intake, largely because snacks weren't photographed. The review lists forgetting as a challenge for every photo method.
• Some foods are harder than others. Foods without a clear shape and foods packed with calories for their size are harder to estimate, and hidden ingredients, the review's example being butter in mashed potatoes, often need a note from the person eating.

Put together, the honest summary is this: a photo can get close on average, especially for simple meals, but any single estimate can be well off, and the meals you don't photograph count for nothing.

What photo apps tend to miss

That cooking oil row matters more than it looks. Two tablespoons in a stir-fry can add over 200 calories that no photo will show, and fat is exactly where the recent test found the biggest misses.

How to get a closer estimate

You don't need to weigh everything. A few habits close most of the gap.

1. Take a better photo. Shoot from above with the whole plate in frame, in good light, with foods spread out rather than piled up. For tall foods like a sandwich or a bowl, add a second photo from the side. A fork or your hand in the frame gives a sense of scale.
2. Check the food names. Fix anything the app got wrong, such as white rice logged as cauliflower rice.
3. Edit the portion. This is the biggest single fix. Use simple guides like cups and spoons, and weigh your most common foods, such as rice, pasta, oats and nut butter, once or twice so you learn what a portion looks like.
4. Add the invisible. Oil, butter, dressing, sugar in your coffee and anything you drink.
5. Photograph everything before you eat, snacks included. A quick photo of a handful of crackers counts as much as dinner does.
6. Judge by the week, not the meal. Single estimates will be off in both directions. What matters is whether your weekly average weight moves the way you expect over two or three weeks. Compare the same week of your cycle each month, since water retention before your period can hide a trend.

If your averages haven't moved after a few weeks of logging, the log is the first thing to check. Our guide to a weight loss plateau walks through portions drifting and other checks, and our guide to a calorie deficit for women explains how to set and adjust a moderate target.

If you'd like to try photo logging, HerShape has it: tap Snap your plate and take a picture of your meal. The habits above apply to any photo log, ours included.

Is photo logging worth it?

For most people, yes, as long as you know what you're getting. Photo logging is fast, and a log you keep up beats a precise one you quit after a week. It's also good at showing patterns, such as how often you eat out, whether breakfast has any protein, or how many vegetables you ate this week.

Where it falls short is precision. If you're eating in a calorie deficit and the app undercounts a few hundred calories a day, your deficit may be much smaller than you think.

For protein, our guide to how much protein women need gives you a daily number to check your log against.

When to skip calorie counting or get support

Calorie counting isn't for everyone. Talk to a clinician or a registered dietitian before you start if you:

• Have a history of disordered eating, or notice that tracking makes you anxious or rigid about food. A dietitian can help you plan with portions and habits instead of numbers.
• Are pregnant or breastfeeding. Your needs change, and eating less isn't the goal.
• Are under 18.
• Have a health condition that changes what or how much you should eat.

If logging ever starts to feel stressful, it's fine to stop. A rough idea of your usual meals is enough for many goals.

The bottom line

You can count calories from a photo, but only roughly. Apps do best with simple plates and struggle with portion size, mixed dishes and anything the camera can't see, and a recent test of four apps found them about a third low on average. Take a clear photo, edit the portion, add oils, sauces and drinks, photograph snacks, and judge your progress by weekly averages over a few weeks, not by any single meal.

Sources

• ScienceDaily (American Society for Nutrition, 2026): Your AI calorie-tracking app may be off by 345 calories
• Annals of Medicine (2023): AI-based digital image dietary assessment methods compared to humans and ground truth: a systematic review
• International Journal of Obesity (2020): Review of the validity and feasibility of image-assisted methods for dietary assessment
• USDA National Agricultural Library: Energy (kcal) per household measure, USDA National Nutrient Database for Standard Reference (PDF)