depth-estimation 0.1.2__tar.gz → 0.1.3__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (101) hide show
  1. {depth_estimation-0.1.2/src/depth_estimation.egg-info → depth_estimation-0.1.3}/PKG-INFO +33 -8
  2. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/README.md +32 -7
  3. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/pyproject.toml +1 -1
  4. {depth_estimation-0.1.2 → depth_estimation-0.1.3/src/depth_estimation.egg-info}/PKG-INFO +33 -8
  5. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/LICENSE +0 -0
  6. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/setup.cfg +0 -0
  7. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/__init__.py +0 -0
  8. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/cli.py +0 -0
  9. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/configuration_utils.py +0 -0
  10. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/__init__.py +0 -0
  11. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/base_dataset.py +0 -0
  12. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/diode.py +0 -0
  13. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/folder.py +0 -0
  14. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/hub.py +0 -0
  15. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/kitti.py +0 -0
  16. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/nyu_depth_v2.py +0 -0
  17. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/data/transforms.py +0 -0
  18. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/evaluation/__init__.py +0 -0
  19. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/evaluation/evaluate.py +0 -0
  20. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/evaluation/metrics.py +0 -0
  21. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/evaluation/profiling.py +0 -0
  22. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/export.py +0 -0
  23. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/losses.py +0 -0
  24. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/modeling_utils.py +0 -0
  25. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/__init__.py +0 -0
  26. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/auto/__init__.py +0 -0
  27. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/auto/modeling_auto.py +0 -0
  28. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/auto/processing_auto.py +0 -0
  29. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v1/__init__.py +0 -0
  30. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v1/configuration_depth_anything_v1.py +0 -0
  31. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v1/modeling_depth_anything_v1.py +0 -0
  32. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v2/__init__.py +0 -0
  33. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v2/configuration_depth_anything_v2.py +0 -0
  34. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v2/modeling_depth_anything_v2.py +0 -0
  35. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v3/__init__.py +0 -0
  36. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v3/configuration_depth_anything_v3.py +0 -0
  37. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_anything_v3/modeling_depth_anything_v3.py +0 -0
  38. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_pro/__init__.py +0 -0
  39. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_pro/configuration_depth_pro.py +0 -0
  40. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/depth_pro/modeling_depth_pro.py +0 -0
  41. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/marigold_dc/__init__.py +0 -0
  42. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/marigold_dc/configuration_marigold_dc.py +0 -0
  43. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/marigold_dc/modeling_marigold_dc.py +0 -0
  44. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/midas/__init__.py +0 -0
  45. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/midas/configuration_midas.py +0 -0
  46. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/midas/modeling_midas.py +0 -0
  47. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/moge/__init__.py +0 -0
  48. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/moge/configuration_moge.py +0 -0
  49. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/moge/modeling_moge.py +0 -0
  50. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/omnivggt/__init__.py +0 -0
  51. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/omnivggt/configuration_omnivggt.py +0 -0
  52. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/omnivggt/modeling_omnivggt.py +0 -0
  53. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/pixel_perfect_depth/__init__.py +0 -0
  54. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/pixel_perfect_depth/configuration_ppd.py +0 -0
  55. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/pixel_perfect_depth/modeling_ppd.py +0 -0
  56. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/vggt/__init__.py +0 -0
  57. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/vggt/configuration_vggt.py +0 -0
  58. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/vggt/modeling_vggt.py +0 -0
  59. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/zoedepth/__init__.py +0 -0
  60. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/zoedepth/configuration_zoedepth.py +0 -0
  61. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/models/zoedepth/modeling_zoedepth.py +0 -0
  62. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/output.py +0 -0
  63. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/pipeline_utils.py +0 -0
  64. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/pipelines/__init__.py +0 -0
  65. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/pipelines/depth_estimation.py +0 -0
  66. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/processing_utils.py +0 -0
  67. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/pruning.py +0 -0
  68. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/quantization.py +0 -0
  69. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/registry.py +0 -0
  70. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/trainer.py +0 -0
  71. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/training_args.py +0 -0
  72. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/video.py +0 -0
  73. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/viz/__init__.py +0 -0
  74. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/viz/animate.py +0 -0
  75. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/viz/composite.py +0 -0
  76. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/viz/display.py +0 -0
  77. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation/viz/error.py +0 -0
  78. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation.egg-info/SOURCES.txt +0 -0
  79. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation.egg-info/dependency_links.txt +0 -0
  80. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation.egg-info/entry_points.txt +0 -0
  81. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation.egg-info/requires.txt +0 -0
  82. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/src/depth_estimation.egg-info/top_level.txt +0 -0
  83. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_auto.py +0 -0
  84. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_cli.py +0 -0
  85. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_config.py +0 -0
  86. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_data.py +0 -0
  87. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_evaluation.py +0 -0
  88. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_export_onnx.py +0 -0
  89. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_inference_all_models.py +0 -0
  90. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_inference_image.py +0 -0
  91. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_losses.py +0 -0
  92. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_model.py +0 -0
  93. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_package.py +0 -0
  94. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_pipeline.py +0 -0
  95. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_processor.py +0 -0
  96. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_pruning.py +0 -0
  97. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_quantization.py +0 -0
  98. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_training.py +0 -0
  99. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_transforms.py +0 -0
  100. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_video.py +0 -0
  101. {depth_estimation-0.1.2 → depth_estimation-0.1.3}/tests/test_viz.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: depth_estimation
3
- Version: 0.1.2
3
+ Version: 0.1.3
4
4
  Summary: A Transformers-style Python library for monocular depth estimation — inference, evaluation, and fine-tuning
5
5
  Author: shriarul5273
6
6
  License-Expression: MIT
@@ -58,21 +58,41 @@ Dynamic: license-file
58
58
 
59
59
  <h3 align="center">A unified Python library for monocular depth estimation</h3>
60
60
 
61
- <h3 align="center">Inference · Video & Streaming · Visualization · Fine-Tuning · Evaluation · Dataset Loading</h3>
61
+ <h3 align="center">Inference · Video · Training · Evaluation · ONNX Export · Pruning · Quantization</h3>
62
62
 
63
63
  ---
64
64
 
65
- `depth_estimation` is the model-definition framework for depth estimation. It provides a single, consistent API across **12 model families and 28 variants** — so you can swap models, compare them, and fine-tune them without rewriting your pipeline.
65
+ `depth_estimation` is a unified model framework for monocular depth estimation. It provides one consistent API across **12 model families and 28 variants**, so you can swap models, compare them, fine-tune them, and prepare them for deployment without rewriting your pipeline.
66
66
 
67
- It covers the full workflow end-to-end: run inference with one line, stream depth from video, visualize results, evaluate on standard benchmarks, and fine-tune on custom depth data — all with the same library.
67
+ It covers the workflow end-to-end: run inference with one line, process video, visualize results, evaluate standard benchmarks, fine-tune on custom depth data, optimize models, and export them to ONNX.
68
+
69
+ > **Latest release: v0.1.3 — Documentation Refresh.** It adds clearer installation paths, a documentation index, updated deployment guidance, and corrected ONNX verification dependencies. Read the [release notes](https://github.com/shriarul5273/depth_estimation/blob/main/docs/release_notes.md) or view the [GitHub release](https://github.com/shriarul5273/depth_estimation/releases/tag/v0.1.3).
68
70
 
69
71
  ## Installation
70
72
 
71
73
  ```bash
72
- pip install depth-estimation
74
+ # Core library
75
+ pip install --upgrade depth-estimation
76
+
77
+ # ONNX export
78
+ pip install "depth-estimation[export]"
79
+
80
+ # Export verification and ONNX quantization
81
+ pip install "depth-estimation[export]" onnxruntime
73
82
  ```
74
83
 
75
- See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for optional extras (CUDA, MPS, etc.).
84
+ Python 3.10–3.12 is supported. For GPU execution of exported models, replace `onnxruntime` with `onnxruntime-gpu`. See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for the complete dependency guide.
85
+
86
+ ## Documentation
87
+
88
+ | Goal | Guide |
89
+ |---|---|
90
+ | Choose a model | [Models](https://github.com/shriarul5273/depth_estimation/blob/main/docs/models.md) |
91
+ | Run video or visualize results | [Video](https://github.com/shriarul5273/depth_estimation/blob/main/docs/video.md) · [Visualization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/viz.md) |
92
+ | Load data, evaluate, or fine-tune | [Data](https://github.com/shriarul5273/depth_estimation/blob/main/docs/data.md) · [Evaluation](https://github.com/shriarul5273/depth_estimation/blob/main/docs/evaluation.md) · [Training](https://github.com/shriarul5273/depth_estimation/blob/main/docs/training.md) |
93
+ | Prepare a model for deployment | [ONNX export](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) · [Pruning](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) · [Quantization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/quantization.md) |
94
+ | Use the command line | [CLI reference](https://github.com/shriarul5273/depth_estimation/blob/main/docs/cli.md) |
95
+ | Extend the library | [Adding a model](https://github.com/shriarul5273/depth_estimation/blob/main/docs/adding_a_model.md) |
76
96
 
77
97
  ---
78
98
 
@@ -119,7 +139,7 @@ depth-estimate predict image.jpg --model depth-anything-v2-vitb
119
139
  Switch from Depth Anything to DepthPro to MoGe by changing a single string. Preprocessing, postprocessing, and output format are identical across all models.
120
140
 
121
141
  **2. The full depth workflow in one place.**
122
- Most libraries stop at inference. This one covers training, evaluation on standard benchmarks, and dataset loading — so you don't have to stitch together separate tools.
142
+ Most libraries stop at inference. This one also covers video, training, benchmark evaluation, dataset loading, pruning, quantization, and ONNX export, so you don't have to stitch together separate tools.
123
143
 
124
144
  **3. Modular, single-file model design.**
125
145
  Each model lives in one self-contained file. No hidden abstractions. If you need to understand or modify a model, there's exactly one place to look. New models self-register — `AutoDepthModel` and `pipeline()` resolve them automatically.
@@ -127,6 +147,9 @@ Each model lives in one self-contained file. No hidden abstractions. If you need
127
147
  **4. Designed for research.**
128
148
  Trainable models with backbone freeze schedules, proper batch-level metric accumulation (no mean-of-means), and a `compare()` function that shows a formatted table across models.
129
149
 
150
+ **5. Built for reliable deployment.**
151
+ Export verification checks ONNX output against PyTorch, quantization verifies accuracy by default, and unsupported model paths fail early with actionable errors.
152
+
130
153
  ---
131
154
 
132
155
  ## Supported Models
@@ -320,7 +343,7 @@ export_onnx(model, "depth_anything_v2_vitb.onnx", input_size=518, verify=True)
320
343
  depth-estimate export --model depth-anything-v2-vitb --output model.onnx --verify
321
344
  ```
322
345
 
323
- Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`. `zoedepth` and `marigold-dc` are **not exportable** (both wrap an opaque external pipeline) — `export_onnx()` raises a clear error immediately rather than a doomed attempt. Requires the optional `onnx` package: `pip install "depth-estimation[export]"`. GPU inference on the exported file works too via `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table, known limitations, and GPU setup.
346
+ Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`, with model-specific limitations documented in the compatibility table. `zoedepth` and `marigold-dc` are **not exportable** because both wrap opaque external pipelines; `export_onnx()` fails early with a clear error. Export requires `pip install "depth-estimation[export]"`; using `verify=True` additionally requires `onnxruntime` or `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table and GPU setup.
324
347
 
325
348
  </details>
326
349
 
@@ -339,6 +362,8 @@ model.export_onnx("pruned.onnx", verify=True) # pruned models export like any o
339
362
 
340
363
  Pure PyTorch (`torch.nn.utils.prune`) — no special hardware or SDK. See [docs/pruning.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) for prune-aware fine-tuning and what pruning does/doesn't buy you (it zeros weights, it doesn't shrink tensors).
341
364
 
365
+ For a complete prune → fine-tune → export → quantize workflow, see [`examples/optimize.py`](https://github.com/shriarul5273/depth_estimation/blob/main/examples/optimize.py).
366
+
342
367
  </details>
343
368
 
344
369
  <details>
@@ -11,21 +11,41 @@
11
11
 
12
12
  <h3 align="center">A unified Python library for monocular depth estimation</h3>
13
13
 
14
- <h3 align="center">Inference · Video & Streaming · Visualization · Fine-Tuning · Evaluation · Dataset Loading</h3>
14
+ <h3 align="center">Inference · Video · Training · Evaluation · ONNX Export · Pruning · Quantization</h3>
15
15
 
16
16
  ---
17
17
 
18
- `depth_estimation` is the model-definition framework for depth estimation. It provides a single, consistent API across **12 model families and 28 variants** — so you can swap models, compare them, and fine-tune them without rewriting your pipeline.
18
+ `depth_estimation` is a unified model framework for monocular depth estimation. It provides one consistent API across **12 model families and 28 variants**, so you can swap models, compare them, fine-tune them, and prepare them for deployment without rewriting your pipeline.
19
19
 
20
- It covers the full workflow end-to-end: run inference with one line, stream depth from video, visualize results, evaluate on standard benchmarks, and fine-tune on custom depth data — all with the same library.
20
+ It covers the workflow end-to-end: run inference with one line, process video, visualize results, evaluate standard benchmarks, fine-tune on custom depth data, optimize models, and export them to ONNX.
21
+
22
+ > **Latest release: v0.1.3 — Documentation Refresh.** It adds clearer installation paths, a documentation index, updated deployment guidance, and corrected ONNX verification dependencies. Read the [release notes](https://github.com/shriarul5273/depth_estimation/blob/main/docs/release_notes.md) or view the [GitHub release](https://github.com/shriarul5273/depth_estimation/releases/tag/v0.1.3).
21
23
 
22
24
  ## Installation
23
25
 
24
26
  ```bash
25
- pip install depth-estimation
27
+ # Core library
28
+ pip install --upgrade depth-estimation
29
+
30
+ # ONNX export
31
+ pip install "depth-estimation[export]"
32
+
33
+ # Export verification and ONNX quantization
34
+ pip install "depth-estimation[export]" onnxruntime
26
35
  ```
27
36
 
28
- See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for optional extras (CUDA, MPS, etc.).
37
+ Python 3.10–3.12 is supported. For GPU execution of exported models, replace `onnxruntime` with `onnxruntime-gpu`. See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for the complete dependency guide.
38
+
39
+ ## Documentation
40
+
41
+ | Goal | Guide |
42
+ |---|---|
43
+ | Choose a model | [Models](https://github.com/shriarul5273/depth_estimation/blob/main/docs/models.md) |
44
+ | Run video or visualize results | [Video](https://github.com/shriarul5273/depth_estimation/blob/main/docs/video.md) · [Visualization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/viz.md) |
45
+ | Load data, evaluate, or fine-tune | [Data](https://github.com/shriarul5273/depth_estimation/blob/main/docs/data.md) · [Evaluation](https://github.com/shriarul5273/depth_estimation/blob/main/docs/evaluation.md) · [Training](https://github.com/shriarul5273/depth_estimation/blob/main/docs/training.md) |
46
+ | Prepare a model for deployment | [ONNX export](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) · [Pruning](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) · [Quantization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/quantization.md) |
47
+ | Use the command line | [CLI reference](https://github.com/shriarul5273/depth_estimation/blob/main/docs/cli.md) |
48
+ | Extend the library | [Adding a model](https://github.com/shriarul5273/depth_estimation/blob/main/docs/adding_a_model.md) |
29
49
 
30
50
  ---
31
51
 
@@ -72,7 +92,7 @@ depth-estimate predict image.jpg --model depth-anything-v2-vitb
72
92
  Switch from Depth Anything to DepthPro to MoGe by changing a single string. Preprocessing, postprocessing, and output format are identical across all models.
73
93
 
74
94
  **2. The full depth workflow in one place.**
75
- Most libraries stop at inference. This one covers training, evaluation on standard benchmarks, and dataset loading — so you don't have to stitch together separate tools.
95
+ Most libraries stop at inference. This one also covers video, training, benchmark evaluation, dataset loading, pruning, quantization, and ONNX export, so you don't have to stitch together separate tools.
76
96
 
77
97
  **3. Modular, single-file model design.**
78
98
  Each model lives in one self-contained file. No hidden abstractions. If you need to understand or modify a model, there's exactly one place to look. New models self-register — `AutoDepthModel` and `pipeline()` resolve them automatically.
@@ -80,6 +100,9 @@ Each model lives in one self-contained file. No hidden abstractions. If you need
80
100
  **4. Designed for research.**
81
101
  Trainable models with backbone freeze schedules, proper batch-level metric accumulation (no mean-of-means), and a `compare()` function that shows a formatted table across models.
82
102
 
103
+ **5. Built for reliable deployment.**
104
+ Export verification checks ONNX output against PyTorch, quantization verifies accuracy by default, and unsupported model paths fail early with actionable errors.
105
+
83
106
  ---
84
107
 
85
108
  ## Supported Models
@@ -273,7 +296,7 @@ export_onnx(model, "depth_anything_v2_vitb.onnx", input_size=518, verify=True)
273
296
  depth-estimate export --model depth-anything-v2-vitb --output model.onnx --verify
274
297
  ```
275
298
 
276
- Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`. `zoedepth` and `marigold-dc` are **not exportable** (both wrap an opaque external pipeline) — `export_onnx()` raises a clear error immediately rather than a doomed attempt. Requires the optional `onnx` package: `pip install "depth-estimation[export]"`. GPU inference on the exported file works too via `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table, known limitations, and GPU setup.
299
+ Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`, with model-specific limitations documented in the compatibility table. `zoedepth` and `marigold-dc` are **not exportable** because both wrap opaque external pipelines; `export_onnx()` fails early with a clear error. Export requires `pip install "depth-estimation[export]"`; using `verify=True` additionally requires `onnxruntime` or `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table and GPU setup.
277
300
 
278
301
  </details>
279
302
 
@@ -292,6 +315,8 @@ model.export_onnx("pruned.onnx", verify=True) # pruned models export like any o
292
315
 
293
316
  Pure PyTorch (`torch.nn.utils.prune`) — no special hardware or SDK. See [docs/pruning.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) for prune-aware fine-tuning and what pruning does/doesn't buy you (it zeros weights, it doesn't shrink tensors).
294
317
 
318
+ For a complete prune → fine-tune → export → quantize workflow, see [`examples/optimize.py`](https://github.com/shriarul5273/depth_estimation/blob/main/examples/optimize.py).
319
+
295
320
  </details>
296
321
 
297
322
  <details>
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
4
 
5
5
  [project]
6
6
  name = "depth_estimation"
7
- version = "0.1.2"
7
+ version = "0.1.3"
8
8
  authors = [
9
9
  { name = "shriarul5273" },
10
10
  ]
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: depth_estimation
3
- Version: 0.1.2
3
+ Version: 0.1.3
4
4
  Summary: A Transformers-style Python library for monocular depth estimation — inference, evaluation, and fine-tuning
5
5
  Author: shriarul5273
6
6
  License-Expression: MIT
@@ -58,21 +58,41 @@ Dynamic: license-file
58
58
 
59
59
  <h3 align="center">A unified Python library for monocular depth estimation</h3>
60
60
 
61
- <h3 align="center">Inference · Video & Streaming · Visualization · Fine-Tuning · Evaluation · Dataset Loading</h3>
61
+ <h3 align="center">Inference · Video · Training · Evaluation · ONNX Export · Pruning · Quantization</h3>
62
62
 
63
63
  ---
64
64
 
65
- `depth_estimation` is the model-definition framework for depth estimation. It provides a single, consistent API across **12 model families and 28 variants** — so you can swap models, compare them, and fine-tune them without rewriting your pipeline.
65
+ `depth_estimation` is a unified model framework for monocular depth estimation. It provides one consistent API across **12 model families and 28 variants**, so you can swap models, compare them, fine-tune them, and prepare them for deployment without rewriting your pipeline.
66
66
 
67
- It covers the full workflow end-to-end: run inference with one line, stream depth from video, visualize results, evaluate on standard benchmarks, and fine-tune on custom depth data — all with the same library.
67
+ It covers the workflow end-to-end: run inference with one line, process video, visualize results, evaluate standard benchmarks, fine-tune on custom depth data, optimize models, and export them to ONNX.
68
+
69
+ > **Latest release: v0.1.3 — Documentation Refresh.** It adds clearer installation paths, a documentation index, updated deployment guidance, and corrected ONNX verification dependencies. Read the [release notes](https://github.com/shriarul5273/depth_estimation/blob/main/docs/release_notes.md) or view the [GitHub release](https://github.com/shriarul5273/depth_estimation/releases/tag/v0.1.3).
68
70
 
69
71
  ## Installation
70
72
 
71
73
  ```bash
72
- pip install depth-estimation
74
+ # Core library
75
+ pip install --upgrade depth-estimation
76
+
77
+ # ONNX export
78
+ pip install "depth-estimation[export]"
79
+
80
+ # Export verification and ONNX quantization
81
+ pip install "depth-estimation[export]" onnxruntime
73
82
  ```
74
83
 
75
- See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for optional extras (CUDA, MPS, etc.).
84
+ Python 3.10–3.12 is supported. For GPU execution of exported models, replace `onnxruntime` with `onnxruntime-gpu`. See [docs/dependencies.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/dependencies.md) for the complete dependency guide.
85
+
86
+ ## Documentation
87
+
88
+ | Goal | Guide |
89
+ |---|---|
90
+ | Choose a model | [Models](https://github.com/shriarul5273/depth_estimation/blob/main/docs/models.md) |
91
+ | Run video or visualize results | [Video](https://github.com/shriarul5273/depth_estimation/blob/main/docs/video.md) · [Visualization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/viz.md) |
92
+ | Load data, evaluate, or fine-tune | [Data](https://github.com/shriarul5273/depth_estimation/blob/main/docs/data.md) · [Evaluation](https://github.com/shriarul5273/depth_estimation/blob/main/docs/evaluation.md) · [Training](https://github.com/shriarul5273/depth_estimation/blob/main/docs/training.md) |
93
+ | Prepare a model for deployment | [ONNX export](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) · [Pruning](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) · [Quantization](https://github.com/shriarul5273/depth_estimation/blob/main/docs/quantization.md) |
94
+ | Use the command line | [CLI reference](https://github.com/shriarul5273/depth_estimation/blob/main/docs/cli.md) |
95
+ | Extend the library | [Adding a model](https://github.com/shriarul5273/depth_estimation/blob/main/docs/adding_a_model.md) |
76
96
 
77
97
  ---
78
98
 
@@ -119,7 +139,7 @@ depth-estimate predict image.jpg --model depth-anything-v2-vitb
119
139
  Switch from Depth Anything to DepthPro to MoGe by changing a single string. Preprocessing, postprocessing, and output format are identical across all models.
120
140
 
121
141
  **2. The full depth workflow in one place.**
122
- Most libraries stop at inference. This one covers training, evaluation on standard benchmarks, and dataset loading — so you don't have to stitch together separate tools.
142
+ Most libraries stop at inference. This one also covers video, training, benchmark evaluation, dataset loading, pruning, quantization, and ONNX export, so you don't have to stitch together separate tools.
123
143
 
124
144
  **3. Modular, single-file model design.**
125
145
  Each model lives in one self-contained file. No hidden abstractions. If you need to understand or modify a model, there's exactly one place to look. New models self-register — `AutoDepthModel` and `pipeline()` resolve them automatically.
@@ -127,6 +147,9 @@ Each model lives in one self-contained file. No hidden abstractions. If you need
127
147
  **4. Designed for research.**
128
148
  Trainable models with backbone freeze schedules, proper batch-level metric accumulation (no mean-of-means), and a `compare()` function that shows a formatted table across models.
129
149
 
150
+ **5. Built for reliable deployment.**
151
+ Export verification checks ONNX output against PyTorch, quantization verifies accuracy by default, and unsupported model paths fail early with actionable errors.
152
+
130
153
  ---
131
154
 
132
155
  ## Supported Models
@@ -320,7 +343,7 @@ export_onnx(model, "depth_anything_v2_vitb.onnx", input_size=518, verify=True)
320
343
  depth-estimate export --model depth-anything-v2-vitb --output model.onnx --verify
321
344
  ```
322
345
 
323
- Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`. `zoedepth` and `marigold-dc` are **not exportable** (both wrap an opaque external pipeline) — `export_onnx()` raises a clear error immediately rather than a doomed attempt. Requires the optional `onnx` package: `pip install "depth-estimation[export]"`. GPU inference on the exported file works too via `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table, known limitations, and GPU setup.
346
+ Verified working for `depth-anything-v1/v2/v3`, `depth-pro`, and `moge`, with model-specific limitations documented in the compatibility table. `zoedepth` and `marigold-dc` are **not exportable** because both wrap opaque external pipelines; `export_onnx()` fails early with a clear error. Export requires `pip install "depth-estimation[export]"`; using `verify=True` additionally requires `onnxruntime` or `onnxruntime-gpu`. See [docs/export.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/export.md) for the full supported-models table and GPU setup.
324
347
 
325
348
  </details>
326
349
 
@@ -339,6 +362,8 @@ model.export_onnx("pruned.onnx", verify=True) # pruned models export like any o
339
362
 
340
363
  Pure PyTorch (`torch.nn.utils.prune`) — no special hardware or SDK. See [docs/pruning.md](https://github.com/shriarul5273/depth_estimation/blob/main/docs/pruning.md) for prune-aware fine-tuning and what pruning does/doesn't buy you (it zeros weights, it doesn't shrink tensors).
341
364
 
365
+ For a complete prune → fine-tune → export → quantize workflow, see [`examples/optimize.py`](https://github.com/shriarul5273/depth_estimation/blob/main/examples/optimize.py).
366
+
342
367
  </details>
343
368
 
344
369
  <details>