Add verified D-FINE-N COCO ONNX bundle

LICENSES.md

@@ -1,6 +1,6 @@
 # Model licenses and attribution
 
-All sixteen bundles are distributed under the Apache License 2.0 terms of
+All seventeen bundles are distributed under the Apache License 2.0 terms of
 their respective upstream projects. Source and transform revisions are recorded
 in `MANIFEST.json`; every SAM 2, SAM 2.1, and EfficientViT-SAM archive also
 contains the applicable complete upstream license text.
@@ -12,6 +12,7 @@ contains the applicable complete upstream license text.
 | SAM decoder used by MobileSAM | [facebookresearch/segment-anything](https://github.com/facebookresearch/segment-anything) | [Apache License 2.0](https://github.com/facebookresearch/segment-anything/blob/main/LICENSE) |
 | EfficientViT-SAM L0/L1/L2/XL0/XL1 | [mit-han-lab/efficientvit](https://github.com/mit-han-lab/efficientvit/tree/de7d7733cc0329f391b33f1f459271562ec27bd5/applications/efficientvit_sam) | [Apache License 2.0](https://github.com/mit-han-lab/efficientvit/blob/de7d7733cc0329f391b33f1f459271562ec27bd5/LICENSE) |
 | RF-DETR Nano detection and Segmentation Nano | [roboflow/rf-detr](https://github.com/roboflow/rf-detr/tree/9b009fa928d6218320439803d1da01869a85c072) | [Apache License 2.0](https://github.com/roboflow/rf-detr/blob/9b009fa928d6218320439803d1da01869a85c072/LICENSE) |
+| D-FINE-N COCO-only detection | [Peterande/D-FINE](https://github.com/Peterande/D-FINE/tree/956d1709314c2c6a4df6f34de232054578a7449f) | [Apache License 2.0](https://github.com/Peterande/D-FINE/blob/956d1709314c2c6a4df6f34de232054578a7449f/LICENSE); [maintainer confirmation for COCO-only checkpoints](https://github.com/Peterande/D-FINE/issues/357#issuecomment-5344034723) |
 
 Copyright remains with the original authors and contributors. AnyLearning and
 Neural Research Lab do not claim ownership of the underlying model research or

MANIFEST.json

@@ -2,6 +2,67 @@
   "schema_version": 1,
   "created_at": "2026-08-31",
   "files": {
+    "dfine_n_coco_956d170.zip": {
+      "archive_members": [
+        "CHECKSUMS.sha256",
+        "EXPORT_REPORT.json",
+        "LICENSE",
+        "PROVENANCE.json",
+        "README.md",
+        "dfine_n_coco.onnx",
+        "export_checked.py"
+      ],
+      "members": {
+        "CHECKSUMS.sha256": {
+          "sha256": "598a7dbfd9e384dce24aaab326c94c8bb77b9c287d0aa2fb11c59c7af0b6247e",
+          "size_bytes": 485
+        },
+        "EXPORT_REPORT.json": {
+          "sha256": "5f31523c80558bb9dccb9cabd32a050365772fedc447244e2adb3a7ef9db1e81",
+          "size_bytes": 4945
+        },
+        "LICENSE": {
+          "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
+          "size_bytes": 11357
+        },
+        "PROVENANCE.json": {
+          "sha256": "bd15d4242ba4aa08f39366d0c0189cc620173dea4ebccf17e0d27c939de29e0a",
+          "size_bytes": 2796
+        },
+        "README.md": {
+          "sha256": "2906ac4928b1f5ef57c55c73a787394191206896ff1a86c2e5d71f21ff523327",
+          "size_bytes": 1160
+        },
+        "dfine_n_coco.onnx": {
+          "sha256": "47593baffd2b340a124939356515043edae9ba0d1cd59473146349ab7a87d832",
+          "size_bytes": 15557424
+        },
+        "export_checked.py": {
+          "sha256": "d16b6ad1c22675b419797b4c89b06f06bf466f1a0bf2cbee148fef707375cfb5",
+          "size_bytes": 9025
+        }
+      },
+      "sha256": "f753f6e552632ef1696ec53c92ccb1ca374dbf2f014c6e4ab005ab98efa3a6e8",
+      "size_bytes": 13979719,
+      "source_checkpoint_sha256": "41973938d2784d38a9836990d805b8392855ebf611aba55f0f7add90e110744c",
+      "source_checkpoint_url": "https://github.com/Peterande/storage/releases/download/dfinev1.0/dfine_n_coco.pth",
+      "source_repo": "Peterande/D-FINE",
+      "source_revision": "956d1709314c2c6a4df6f34de232054578a7449f",
+      "weight_license": "Apache-2.0 (COCO-only checkpoint; Objects365-derived checkpoints excluded)",
+      "weight_license_confirmation": "https://github.com/Peterande/D-FINE/issues/357#issuecomment-5344034723",
+      "export": {
+        "format": "onnx",
+        "opset": 16,
+        "precision": "float32",
+        "static_shape": [1, 3, 640, 640],
+        "checkpoint_load": "weights_only=True, mmap=True"
+      },
+      "parity": {
+        "labels": "exact",
+        "maximum_box_absolute_error_pixels": 0.004058837890625,
+        "maximum_score_absolute_error": 0.000007808208465576172
+      }
+    },
     "rfdetr_nano_detection_1_9_4.zip": {
       "archive_members": [
         "CHECKSUMS.sha256",

README.md

@@ -25,7 +25,7 @@ Versioned ONNX model bundles used by
 [AnyLearning](https://github.com/nrl-ai/anylearning-oss) for local object
 detection, instance segmentation, and prompt-guided image segmentation.
 
-This repository contains the sixteen models currently offered by AnyLearning:
+This repository contains the seventeen models currently offered by AnyLearning:
 
 | File | Model | Download size | SHA-256 |
 | --- | --- | ---: | --- |
@@ -45,11 +45,12 @@ This repository contains the sixteen models currently offered by AnyLearning:
 | `sam2_1_hiera_large.zip` | SAM 2.1 Hiera-Large | 805,293,551 bytes | `15f74c68530b0bc9f37d2189394a100b81541f55362b34f55a309ba12b9e1fa4` |
 | `rfdetr_nano_detection_1_9_4.zip` | RF-DETR Nano detection | 99,769,055 bytes | `5b130a1c2eb01be3bfbda703367b5d993821690de26f4abbb14a94ee3660c5fe` |
 | `rfdetr_nano_segmentation_1_9_4.zip` | RF-DETR Segmentation Nano | 113,602,026 bytes | `133fdb5aed76233a6959addbdb7d64f5132f3cf2b2705a1994c2cbdfc2e93f4d` |
+| `dfine_n_coco_956d170.zip` | D-FINE-N COCO detection | 13,979,719 bytes | `f753f6e552632ef1696ec53c92ccb1ca374dbf2f014c6e4ab005ab98efa3a6e8` |
 
 Each promptable-segmentation ZIP contains a small AnyLearning model
-configuration and one encoder plus one decoder ONNX model. Each RF-DETR ZIP
-contains one static ONNX graph plus its provenance, checksums, and verbatim
-upstream license. `MANIFEST.json` records the exact source revision, archive
+configuration and one encoder plus one decoder ONNX model. Each RF-DETR and
+D-FINE ZIP contains one static ONNX graph plus its provenance, checksums, and
+verbatim upstream license. `MANIFEST.json` records the exact source revision, archive
 size, checksum, and expected members. Every transformed or exported bundle pins
 each extracted member's size and SHA-256 as well as its source artifact
 identity.
@@ -80,6 +81,12 @@ Viet-Anh Nguyen:
   `9b009fa928d6218320439803d1da01869a85c072`. Both graphs use static batch-one
   float32 contracts at opset 17. Native/export parity evidence is recorded
   inside each archive.
+- D-FINE-N detection: the official COCO-only checkpoint and exporter at source
+  revision `956d1709314c2c6a4df6f34de232054578a7449f`. The checkpoint was loaded
+  only by the included restricted weights-only conversion helper. The static
+  opset-16 graph has exact native label parity and less than 0.005-pixel box
+  drift on the retained landscape/portrait corpus. Objects365-derived weights
+  are excluded.
 
 The SAM 2 and SAM 2.1 encoders were transformed by
 [AnyLearning's checksum-gated ONNX-only tool](https://github.com/nrl-ai/anylearning-oss/blob/7575ab52cc0a31e5ec3d71b6ec7157dbe042c7d7/scripts/prepare_sam2_encoder.py).
@@ -95,6 +102,7 @@ Original model projects:
 - [MobileSAM](https://github.com/ChaoningZhang/MobileSAM)
 - [EfficientViT-SAM](https://github.com/mit-han-lab/efficientvit/tree/de7d7733cc0329f391b33f1f459271562ec27bd5/applications/efficientvit_sam)
 - [RF-DETR](https://github.com/roboflow/rf-detr/tree/9b009fa928d6218320439803d1da01869a85c072)
+- [D-FINE](https://github.com/Peterande/D-FINE/tree/956d1709314c2c6a4df6f34de232054578a7449f)
 
 The SAM and MobileSAM files are unchanged mirrors. EfficientViT-SAM encoders
 are unchanged mirrors; their decoders preserve the original learned tensors and
@@ -117,6 +125,8 @@ and editable-shape conversion. These archives are not standalone applications.
 - Intended for interactive point/rectangle-prompt segmentation in AnyLearning.
 - RF-DETR bundles are intended for human-reviewed COCO object detection and
   instance-segmentation suggestions.
+- The D-FINE bundle is intended for human-reviewed COCO object-detection
+  suggestions.
 - Results require human review before becoming dataset labels.
 - Quality and latency vary with image content, hardware, execution provider, and
   model size.
@@ -174,3 +184,6 @@ For EfficientViT-SAM, cite:
 
 For RF-DETR, use the citation requested by the
 [official project](https://github.com/roboflow/rf-detr/tree/9b009fa928d6218320439803d1da01869a85c072).
+
+For D-FINE, use the citation requested by the
+[official project](https://github.com/Peterande/D-FINE/tree/956d1709314c2c6a4df6f34de232054578a7449f).

dfine_n_coco_956d170.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f753f6e552632ef1696ec53c92ccb1ca374dbf2f014c6e4ab005ab98efa3a6e8
+size 13979719