Food Photography Collection

1,200 styled food photographs — overhead and 45°, natural light, with dish and cuisine labels.

A styled food-photography set spanning cuisines, courses and plating styles, shot mostly in soft natural light with clean props. Consistent enough to train a coherent food-photography look, varied enough to cover breakfast to fine dining, with dish and cuisine metadata for retrieval and classification.

Dataset details

Supported tasks

• Food-photography style / LoRA training
• Dish and cuisine classification
• Menu, delivery-app and editorial reference

Dataset structure

Each item ships as a photograph plus dish metadata:

{
  "id": "food_00420",
  "image": "food_00420.jpg",
  "dish": "shakshuka",
  "cuisine": "middle_eastern",
  "course": "breakfast",
  "angle": "overhead",
  "lighting": "soft_natural",
  "props": ["cast_iron", "linen_napkin"]
}

Samples

Motion sample

A short clip rendered with this listing — click to play the HLS stream.

Curation

Dishes were cooked and styled for camera, shot on a small set with controlled natural and diffused light, then labelled by dish, cuisine and course. Duplicates and soft frames were culled.

Personal and sensitive information

None significant — food and tableware only. Occasional hands appear during plating and are cropped or accepted as incidental; no identifiable faces.

Considerations & limitations

• Western and Middle-Eastern cuisines are over-represented relative to others.
• Heavy styling means real-world / homemade food may look off-distribution.
• Prop palette is intentionally muted and repeats across shots.

License & provenance

Cooked, styled and shot by the creator. The sample archive (full_dataset.zip) is a free 50-item preview; the full set of 1,200 items unlocks with a purchase. Redistribution of the raw dataset is not permitted under either license.

Citation

@misc{kumo-studio_00000012,
  title  = {Food Photography Collection},
  author = {Kumo Studio (@kumo-studio)},
  year   = {2026},
  note   = {TensorForge marketplace dataset, LST-00000012-0000-4000-8000-000000000012}
}