Download Now
Methods · References

The formulas and sources behind it

CalorieMate's fat-loss system isn't a marketing story — it's a reusable methodology built on public formulas and national-level nutrition references. Here are the complete sources.

1. Basal Metabolism & Daily Burn (BMR / TDEE)

Mifflin-St Jeor ↗男性 BMR = 10 × 体重(kg) + 6.25 × 身高(cm) − 5 × 年龄 + 5女性 BMR = 10 × 体重(kg) + 6.25 × 身高(cm) − 5 × 年龄 − 161TDEE = BMR × 活动系数

Use:Basal Metabolic Rate (BMR) calculation / Total Daily Energy Expenditure (TDEE) calculation。

Why Mifflin-St Jeor: A widely adopted modern BMR formula recommended by mainstream nutrition science. The older Harris-Benedict formula over-estimates BMR by 5-10%, leading to inflated intake targets and slower-than-expected fat loss.

2. Exercise Burn (MET — Metabolic Equivalent of Task)

MET (Metabolic Equivalent of Task)
消耗(kcal) = MET × 体重(kg) × 时长(小时)

Coverage: 600+ types (running / strength / yoga / swimming / cycling / hiking / jump rope / HIIT / elliptical / spin / dance / skiing / skating / badminton / tennis / basketball / soccer / table tennis / pull-ups / squats / crunches and more).

Use:Calorie burn estimation for 600+ exercise types. Low / medium / high intensity adjustments closely match reality.

3. Food Database References

Our 500K-entry Chinese food database is built and maintained by the dietitian team, strictly drawing on the following 4 national-level references:

4. AI Companion Model Architecture

Base model
DeepSeek deep-reasoning model
Retrieval augmentation
RAG (Retrieval-Augmented Generation)
Knowledge base
In-house nutrition & exercise knowledge base covering dietary guidelines, chronic-disease diets, pregnancy nutrition, low-GI, keto, low-FODMAP, and more
Persona system
Caring friend / tough-love coach / professional dietitian / custom

⚠️ AI responses are for general dietary and exercise reference only and do not constitute medical advice.

5. Photo Recognition Technical Layers

  1. Layer 1 · Vision LLM (VLM) for food understandingA general vision-language model identifies the ingredients, cooking method, condiments, and container type
  2. Layer 2 · 500K-entry Chinese food database for precise nutritional matchingCalibrates the recognition output against our in-house Chinese food database
  3. Layer 3 · AR portion ranging for volume & weight estimationCombines AR depth sensing, plane detection, and tableware reference ratios to estimate food volume and weight
← Team credentials & university researchAwards & media recognition →