CalPulse
Snap the Menu. Get Instant Calories & Macros
608 followers
Snap the Menu. Get Instant Calories & Macros
608 followers
Snap any menu and instantly get calories, macros & healthier swaps. Make informed choices while deciding, not regretting. Perfect for dining out, travel, and food delivery.
Payment Required
Launch Team / Built With
PawChamp
Congrats and good luck guys. You are really great. Proud to share the platform today with you 🙌
CalPulse
@loniuk Thank you, Nik! The feeling is mutual. Honored to be launching alongside you and the PawChamp team. Let's both have a great day! 🚀
CalPulse
@loniuk Thank you so much! Your support means a lot to us. We're proud to be on this journey together and appreciate your encouragement. 💪
I think it's a very good idea. Based on the photos I see, the app gives a definite amount rather than an estimate.
Since the AI has no way of knowing the quantity of the food, as well as factors such as whether they used lean or fatty meat, how much oil and cream they used, etc, what is the accuracy like?
CalPulse
@footer Thanks for your support!🙌
To tackle the accuracy challenge, CalPulse uses a multi-model inference pipeline:
OCR models for understanding text layout
Food vision models to distinguish between similar dishes
Localized nutrition databases to adapt to regional variations
This setup helps us deliver real-time, context-aware nutrition analysis, even considering factors like ingredient variations.
@peilan_qin Sure, but that still doesn't address the issue. If the app tells me a food is 600c, that is certainly inaccurate. Regardless of the technologies used the app has no way of knowing how fatty the meat is, or even if it's pork or meat. It doensn't know whe kind of cheese used, etc. So a logical approach would be to provide an estimate, like 600-800c.
首先，我想说的是，恭喜发布，然后，嗯，有一个问题：如何根据菜单计算这份餐品的实际卡路里。我当然知道总的卡路里等于每种配菜的单位卡路里乘以配菜的量，再把各种配菜的卡路里总量加起来就是这道餐的总卡路里量，但是，怎么通过菜单判断这道菜里面配菜的量是多少？比如，一个汉堡的菜单图片，怎么在汉堡交到你手上之前计算这个汉堡实际有多大？怎么根据番茄炒蛋的图片，计算番茄炒蛋这道菜用了多少鸡蛋，多少番茄，以及多少植物油？个人感觉这个误差会非常的大
CalPulse
@peter_wang20 Great question. You're right, exact measurement from a menu is impossible. Our AI estimates portions by comparing dishes to thousands of labeled food images in our database. We focus on providing accurate comparisons (e.g. "this burger has 25% more calories than that salad") rather than absolute precision. This gives you the actionable insight needed to make smarter choices.
CalPulse
@peter_wang20 thanks! this is a good question and here is something I'd like to share:
there’s no such thing as a “standard meal.”, we fine-tuned several vision-language models on food recognition datasets, combining them with multiple regional food databases. Each model has its own strength, which means a dynamic routing system will compare outputs in real time.
To ensure precision, we applied a reinforcement-learning technology to adjust model weights. Over time, the AI learns that “one plate of nasi goreng” in Jakarta doesn’t equal the same calories as in Singapore.
CalPulse
👨💻 Engineer here from the CalPulse dev team.
This project started as a late-night experiment: “Can we make AI understand a restaurant menu like a human?”
Turns out… not that easy 😅
Menus mix fonts, layouts, slang, and even emojis — so a single model couldn’t handle it. We built a multi-model pipeline that routes between OCR, food vision models, and regional nutrition databases in real time.
Each model has strengths (ingredient detection, portion reasoning, localization), and the system reinforces itself through user feedback — kind of like a small RL loop for food 🍜.
Now when someone snaps a menu and gets accurate, localized results in seconds — that’s the moment all the late-night debugging feels worth it.
Proud of the team. Still improving every meal. 🚀
CalPulse
@tonyzhu Can confirm 😂 Tony’s the one still in the office when everyone else’s gone home.
Pretty sure the AI isn’t the only thing learning overnight — it’s him too.
Theysaid
I really like the idea of snapping a menu and instantly getting calories, macros and healthier-swap options.
CalPulse
@chrishicken So glad this clicks with you! We've all been there - menu in hand, trying to make sense of options. That's why we obsessed over making this as frictionless as possible.
CalPulse
@chrishicken Thank you! We’re happy you like the idea. 😃
CalPulse
As part of the CalPulse dev team, I’m really proud of how this idea turned into something real.
Our mission was simple — make healthy eating effortless anywhere in the world — but turning a single menu photo into accurate calorie and budget recommendations was a long, messy journey.
Early on, we learned there’s no such thing as a “standard meal.” Menus rarely mention portions or nutrition facts, so we couldn’t rely on a single AI model. Instead, we fine-tuned several vision-language models on food recognition datasets, combining them with multiple regional food databases. Each model has its own strength — one better at ingredient detection, another at portion reasoning — so we built a dynamic routing system that compares outputs in real time.
To ensure precision, the final prediction isn’t just averaged — it’s reinforced. We applied a reinforcement-learning loop where real-world user feedback and known nutrition data continuously adjust model weights. Over time, the AI learns that “one plate of nasi goreng” in Jakarta doesn’t equal the same calories as in Singapore.
There were countless nights of debugging when even rice bowls confused the models 😅, but that process taught us where machine perception meets human behavior.
Now, seeing people snap a menu and instantly get accurate, localized meal insights makes all the late-night experiments worth it. CalPulse is still evolving — and still learning from every plate. 🚀
CalPulse
@forrest_chen_fk I'm so proud of our team's dedication and innovation.
Here’s to continuous learning and improving CalPulse together! 💪
CalPulse
@forrest_chen_fk Haha, Forrest basically lived in front of the training monitor for weeks — I’m still not sure if he eats or just fine-tunes himself now 😂
But seriously, couldn’t have pulled this off without him. The “rice bowl confusion” era was wild 🍚💪
CalPulse
As part of the marketing team for CalPulse, I’m super excited to see our vision come to life! 😊
Our goal is all about making healthy eating easy for everyone, and it’s amazing to see how our work is changing the way people enjoy their meals.
I love being part of the team that connects with users and shares their stories. Every piece of feedback helps us get better and better. We’re not just sharing insights, we’re building a community focused on wellness and convenience.
Here’s to more awesome milestones ahead! 💪
CalPulse
@peilan_qin While the rest of us debug calories, Peilan debugs emotions 😄
Seriously though, love how she brings heart to the tech — couldn’t have done it without you! 🙌