DepthPro
by apple
Zero-shot metric monocular depth estimation with sharp boundaries in under a second
apple/DepthPromixpeek://image_extractor@v1/apple_depthpro_v1Overview
DepthPro is Apple's foundation model for zero-shot metric monocular depth estimation, producing 2.25-megapixel depth maps (1536x1536) in 0.3 seconds on a V100 GPU. Unlike relative depth models, DepthPro predicts absolute metric depth without requiring camera intrinsics, and includes a built-in focal length estimator. Its multi-scale ViT architecture with a shared DINOv2 encoder and DPT-like fusion stage preserves sharp object boundaries.
On Mixpeek, DepthPro enables metric-accurate spatial understanding of images and video frames, powering use cases like 3D scene reconstruction, spatial filtering in retrieval, and depth-aware content organization.
Architecture
Multi-scale Vision Transformer with shared DINOv2 encoder processing image patches at multiple resolutions. DPT-like fusion stage merges and upsamples features for dense prediction. Built-in focal length estimation head. Outputs 1536x1536 metric depth maps with absolute scale.
Mixpeek SDK Integration
// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so DepthPro runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
"https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
{
method: "POST",
headers: {
Authorization: "Bearer API_KEY",
"Content-Type": "application/json",
},
body: JSON.stringify({
collection_id: "col_your_collection",
documents: [
{
document_id: "asset-00412",
// Boxes, masks, depth maps and anomaly scores are structured
// results, not vectors. They go in payload and are reachable
// through pre_filters on a retriever, not through similarity.
payload: {
detections: modelOutput,
source_key: "archive/2026/asset-00412",
},
},
],
}),
},
);
// No managed alternative for an open label set. Two extractors do emit a
// bbox, for the one thing each detects: document_graph_extractor@v1 per
// layout block, face_identity_extractor@v1 per face. Nothing ships that
// returns masks, depth maps or anomaly scores.Capabilities
- Zero-shot metric depth (absolute scale, no camera intrinsics needed)
- 2.25-megapixel output (1536x1536) in 0.3s
- Sharp boundary preservation via multi-scale architecture
- Built-in focal length estimation from a single image
- State-of-the-art boundary accuracy metrics
Use Cases on Mixpeek
Benchmarks
| Dataset | Metric | Score | Source |
|---|---|---|---|
| NYUv2 | AbsRel | 0.036 | Bochkovskii et al., 2024: Depth Pro paper |
| KITTI | AbsRel | 0.039 | Bochkovskii et al., 2024: Depth Pro paper |
| Boundary F1 | F1 (depth edges) | State-of-the-art | Bochkovskii et al., 2024: Depth Pro paper |
Performance
Common Pipeline Companions
Explore on Mixpeek
Compare alternatives in this category
Hand-picked tools & platforms compared
See how Mixpeek runs models as extractors
Store & search embeddings at scale
Usage-based pricing for pipelines
Compare models, APIs & infrastructure
Specification
Research Paper
Depth Pro: Sharp Monocular Metric Depth in Less Than a Second
arxiv.orgBuild a pipeline with DepthPro
Add this model to a processing pipeline alongside other extractors. Combine with retrieval stages for end-to-end search.
Run it on your own data, free