Rename model to Image2SceneGraph
Browse files- .gitattributes +1 -1
- INFERENCE.md +1 -1
- NOTICE.md +2 -2
- OUTPUT_FORMAT.md +1 -1
- README.md +9 -9
- SHA256SUMS +8 -8
- assets/{image2json-results.png → image2scenegraph-results.png} +0 -0
- release_metadata.json +1 -1
- smoke/validation.json +1 -1
.gitattributes
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 2 |
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 3 |
-
assets/
|
|
|
|
| 1 |
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 2 |
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
assets/image2scenegraph-results.png filter=lfs diff=lfs merge=lfs -text
|
INFERENCE.md
CHANGED
|
@@ -14,7 +14,7 @@ import os
|
|
| 14 |
from pathlib import Path
|
| 15 |
|
| 16 |
def main():
|
| 17 |
-
model_dir = Path("./
|
| 18 |
image_path = Path("./example.png")
|
| 19 |
spec = json.loads((model_dir / "inference_spec.json").read_text())
|
| 20 |
for key, value in spec["environment"].items():
|
|
|
|
| 14 |
from pathlib import Path
|
| 15 |
|
| 16 |
def main():
|
| 17 |
+
model_dir = Path("./image2scenegraph-model").resolve()
|
| 18 |
image_path = Path("./example.png")
|
| 19 |
spec = json.loads((model_dir / "inference_spec.json").read_text())
|
| 20 |
for key, value in spec["environment"].items():
|
NOTICE.md
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
# License and attribution
|
| 2 |
|
| 3 |
-
ChartGalaxy++
|
| 4 |
|
| 5 |
The ChartGalaxy++ fine-tuning contributions and newly written documentation are provided under the [Creative Commons Attribution-NonCommercial 4.0 International license](https://creativecommons.org/licenses/by-nc/4.0/). The complete license text is included as [LICENSE](LICENSE).
|
| 6 |
|
| 7 |
Qwen's original model and supporting materials are credited to the **Qwen Team**. Retained upstream materials, including tokenizer and configuration assets, retain their existing license. Qwen publishes its [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) and [Qwen3.5-4B-Base](https://huggingface.co/Qwen/Qwen3.5-4B-Base) releases under Apache-2.0. A copy of the upstream license is included as [LICENSE-QWEN-APACHE-2.0](LICENSE-QWEN-APACHE-2.0). The CC BY-NC notice does not replace the license of independently retained upstream materials.
|
| 8 |
|
| 9 |
-
Attribution for this fine-tuned artifact: **ChartGalaxy++
|
|
|
|
| 1 |
# License and attribution
|
| 2 |
|
| 3 |
+
ChartGalaxy++ Image2SceneGraph is a fine-tuned model in the Qwen3.5-4B family. The model weights were modified through supervised fine-tuning for chart scene graph prediction.
|
| 4 |
|
| 5 |
The ChartGalaxy++ fine-tuning contributions and newly written documentation are provided under the [Creative Commons Attribution-NonCommercial 4.0 International license](https://creativecommons.org/licenses/by-nc/4.0/). The complete license text is included as [LICENSE](LICENSE).
|
| 6 |
|
| 7 |
Qwen's original model and supporting materials are credited to the **Qwen Team**. Retained upstream materials, including tokenizer and configuration assets, retain their existing license. Qwen publishes its [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) and [Qwen3.5-4B-Base](https://huggingface.co/Qwen/Qwen3.5-4B-Base) releases under Apache-2.0. A copy of the upstream license is included as [LICENSE-QWEN-APACHE-2.0](LICENSE-QWEN-APACHE-2.0). The CC BY-NC notice does not replace the license of independently retained upstream materials.
|
| 8 |
|
| 9 |
+
Attribution for this fine-tuned artifact: **ChartGalaxy++ Image2SceneGraph, ChartGalaxyPP**. Retain the license and attribution notices and identify further changes when redistributing modified versions.
|
OUTPUT_FORMAT.md
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
#
|
| 2 |
|
| 3 |
The model emits a JSON object with `high_level_description`, `style_description`, and `elements`. `elements.background` describes the background, and `elements.layout` is a flat list of element and group nodes. A node's `parent` refers to another node label or the implicit `ROOT`.
|
| 4 |
|
|
|
|
| 1 |
+
# Image2SceneGraph output format
|
| 2 |
|
| 3 |
The model emits a JSON object with `high_level_description`, `style_description`, and `elements`. `elements.background` describes the background, and `elements.layout` is a flat list of element and group nodes. A node's `parent` refers to another node label or the implicit `ROOT`.
|
| 4 |
|
README.md
CHANGED
|
@@ -10,21 +10,21 @@ tags:
|
|
| 10 |
- chart
|
| 11 |
- infographic
|
| 12 |
- scene-graph
|
| 13 |
-
-
|
| 14 |
- qwen3_5
|
| 15 |
---
|
| 16 |
|
| 17 |
-
<h1 align="center">ChartGalaxy++ ·
|
| 18 |
|
| 19 |
<p align="center"><strong>From infographic images to structured chart representations</strong></p>
|
| 20 |
|
| 21 |
-
<p align="center"><a href="https://huggingface.co/datasets/ChartGalaxyPP/ChartGalaxyPlusPlus">Dataset</a> · <a href="https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus">Project</a> · <a href="https://huggingface.co/ChartGalaxyPP/ChartGalaxyPlusPlus-
|
| 22 |
|
| 23 |
-
**
|
| 24 |
|
| 25 |
-
 · [Metric definitions](https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus/blob/main/applications/image_to_scene_graph/metric_definitions.json) · [Saved predictions and scores](https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus/releases/download/v1.0/image-to-scene-graph.tar.gz)
|
| 39 |
|
|
@@ -44,9 +44,9 @@ These are the paper's scores, including its output-recovery and graph-evaluation
|
|
| 44 |
```python
|
| 45 |
from huggingface_hub import HfApi, snapshot_download
|
| 46 |
|
| 47 |
-
repo = "ChartGalaxyPP/ChartGalaxyPlusPlus-
|
| 48 |
revision = HfApi().model_info(repo).sha
|
| 49 |
-
snapshot_download(repo, revision=revision, local_dir="
|
| 50 |
```
|
| 51 |
|
| 52 |
Follow **[the complete single-image inference example](INFERENCE.md)** for the prompt, image preprocessing, and generation settings. Keep the weight shards, tokenizer, processor, and chat template together. The tested runtime uses Python 3.11, vLLM 0.20.2, Transformers 5.12.1, and PyTorch 2.11.0 / CUDA 13.0 on one RTX PRO 6000 GPU; exact dependencies are in [requirements-smoke.txt](requirements-smoke.txt).
|
|
|
|
| 10 |
- chart
|
| 11 |
- infographic
|
| 12 |
- scene-graph
|
| 13 |
+
- image2scenegraph
|
| 14 |
- qwen3_5
|
| 15 |
---
|
| 16 |
|
| 17 |
+
<h1 align="center">ChartGalaxy++ · Image2SceneGraph</h1>
|
| 18 |
|
| 19 |
<p align="center"><strong>From infographic images to structured chart representations</strong></p>
|
| 20 |
|
| 21 |
+
<p align="center"><a href="https://huggingface.co/datasets/ChartGalaxyPP/ChartGalaxyPlusPlus">Dataset</a> · <a href="https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus">Project</a> · <a href="https://huggingface.co/ChartGalaxyPP/ChartGalaxyPlusPlus-Image2SceneGraph/blob/main/INFERENCE.md">Run inference</a> · <a href="https://huggingface.co/ChartGalaxyPP/ChartGalaxyPlusPlus-Image2SceneGraph/blob/main/OUTPUT_FORMAT.md">Output format</a></p>
|
| 22 |
|
| 23 |
+
**Image2SceneGraph identifies chart elements and how they belong together.** Fine-tuned from the **Qwen3.5-4B family** on ChartGalaxy++, it predicts visual elements, recognized text, bounding boxes, appearance attributes, semantic groups, and parent links as structured JSON.
|
| 24 |
|
| 25 |
+
|
| 26 |
|
| 27 |
+
*Paper examples: separating nearby visual elements and placing marks in the correct semantic group. [Enlarge](assets/image2scenegraph-results.png)*
|
| 28 |
|
| 29 |
## Results
|
| 30 |
|
|
|
|
| 33 |
| Model | Node F1 | Hierarchy F1 | Spatial F1 |
|
| 34 |
| --- | ---: | ---: | ---: |
|
| 35 |
| GPT-6 Astra | 76.0% | 50.3% | 72.5% |
|
| 36 |
+
| **Image2SceneGraph (ours)** | **89.4%** | **85.1%** | **88.0%** |
|
| 37 |
|
| 38 |
[Full comparison with 10 baselines](https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus/blob/main/applications/image_to_scene_graph/paper_table.json) · [Metric definitions](https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus/blob/main/applications/image_to_scene_graph/metric_definitions.json) · [Saved predictions and scores](https://github.com/ChartGalaxyPP/ChartGalaxyPlusPlus/releases/download/v1.0/image-to-scene-graph.tar.gz)
|
| 39 |
|
|
|
|
| 44 |
```python
|
| 45 |
from huggingface_hub import HfApi, snapshot_download
|
| 46 |
|
| 47 |
+
repo = "ChartGalaxyPP/ChartGalaxyPlusPlus-Image2SceneGraph"
|
| 48 |
revision = HfApi().model_info(repo).sha
|
| 49 |
+
snapshot_download(repo, revision=revision, local_dir="image2scenegraph-model")
|
| 50 |
```
|
| 51 |
|
| 52 |
Follow **[the complete single-image inference example](INFERENCE.md)** for the prompt, image preprocessing, and generation settings. Keep the weight shards, tokenizer, processor, and chat template together. The tested runtime uses Python 3.11, vLLM 0.20.2, Transformers 5.12.1, and PyTorch 2.11.0 / CUDA 13.0 on one RTX PRO 6000 GPU; exact dependencies are in [requirements-smoke.txt](requirements-smoke.txt).
|
SHA256SUMS
CHANGED
|
@@ -1,11 +1,11 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
cc97a2c4ff7723a1f25179dc7c18f676fa84191bbb36ea7d0061acaf3c102559 LICENSE
|
| 4 |
50cbab8a892c5f2993b8c7351a99182507472def3b1374558308605d99b86b32 LICENSE-QWEN-APACHE-2.0
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
30f4be461845fd9c11ac640b5e1ea04b3c72f37523fd5888b353104b5a77cd4a assets/
|
| 9 |
a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715 chat_template.jinja
|
| 10 |
881208c0f3f48d8fe920b4caf1bac23b4e2d351557bf2c1314b2040ceb4c6bba config.json
|
| 11 |
a50023c63e657eff70e2cfdd42f384eedd085ffa34898aa6d7e34cb16618bbc7 inference_spec.json
|
|
@@ -14,10 +14,10 @@ a50023c63e657eff70e2cfdd42f384eedd085ffa34898aa6d7e34cb16618bbc7 inference_spec
|
|
| 14 |
ad380c647e34236cf9f313ff673781e2fca1aaff85af6117727762224f019993 model.safetensors.index.json
|
| 15 |
27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516 preprocessor_config.json
|
| 16 |
d89ef49ce9cd37fbf510158e13c1ef063d9286411c1ec9049932dbe0487143b1 processor_config.json
|
| 17 |
-
|
| 18 |
3906fb04e65a3815d7c4c4248b8ac0e8bfdacbc722ba44491be344ce225ec036 requirements-smoke.txt
|
| 19 |
2510f614f11d9df4393af8822b23386ea4a0ef5323365eb82096d8c3d8947da6 smoke/prediction.json
|
| 20 |
bceee5f5cd9d33d5e9ead0095dfdfa572cbca752c9e35ee696dfcf4b5aad4334 smoke/prediction.raw.txt
|
| 21 |
-
|
| 22 |
06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523 tokenizer.json
|
| 23 |
792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b tokenizer_config.json
|
|
|
|
| 1 |
+
abe4ad1fe803bb42244df4c7631730d0adf79d110764b357b5f0eb16a161ae67 .gitattributes
|
| 2 |
+
edc8cbcf5feea8a86eeb4001cec3609b9788e1b26d7d3cc52d72173849335544 INFERENCE.md
|
| 3 |
cc97a2c4ff7723a1f25179dc7c18f676fa84191bbb36ea7d0061acaf3c102559 LICENSE
|
| 4 |
50cbab8a892c5f2993b8c7351a99182507472def3b1374558308605d99b86b32 LICENSE-QWEN-APACHE-2.0
|
| 5 |
+
77a8b9edad50531d427f38ac66360e25411e0573abe63f502391829d694d195f NOTICE.md
|
| 6 |
+
c36233861503baeac5c5ef541f3ea11e585dbe6e97311ab82c2ac22500754e40 OUTPUT_FORMAT.md
|
| 7 |
+
e787ad2adc00735704c1935355403cfbaecb051924c69c58872de375ccf9be31 README.md
|
| 8 |
+
30f4be461845fd9c11ac640b5e1ea04b3c72f37523fd5888b353104b5a77cd4a assets/image2scenegraph-results.png
|
| 9 |
a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715 chat_template.jinja
|
| 10 |
881208c0f3f48d8fe920b4caf1bac23b4e2d351557bf2c1314b2040ceb4c6bba config.json
|
| 11 |
a50023c63e657eff70e2cfdd42f384eedd085ffa34898aa6d7e34cb16618bbc7 inference_spec.json
|
|
|
|
| 14 |
ad380c647e34236cf9f313ff673781e2fca1aaff85af6117727762224f019993 model.safetensors.index.json
|
| 15 |
27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516 preprocessor_config.json
|
| 16 |
d89ef49ce9cd37fbf510158e13c1ef063d9286411c1ec9049932dbe0487143b1 processor_config.json
|
| 17 |
+
1f63e3f390653b01fa9cf10e32c73b390df1f21eed9ae43e817a4dcd61820f02 release_metadata.json
|
| 18 |
3906fb04e65a3815d7c4c4248b8ac0e8bfdacbc722ba44491be344ce225ec036 requirements-smoke.txt
|
| 19 |
2510f614f11d9df4393af8822b23386ea4a0ef5323365eb82096d8c3d8947da6 smoke/prediction.json
|
| 20 |
bceee5f5cd9d33d5e9ead0095dfdfa572cbca752c9e35ee696dfcf4b5aad4334 smoke/prediction.raw.txt
|
| 21 |
+
ceabb838b3eec7cb5c2666c2f144ae4032602e0cf8d6bf8451a83b91f40c2e3e smoke/validation.json
|
| 22 |
06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523 tokenizer.json
|
| 23 |
792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b tokenizer_config.json
|
assets/{image2json-results.png → image2scenegraph-results.png}
RENAMED
|
File without changes
|
release_metadata.json
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
{"model_name":"ChartGalaxy++
|
|
|
|
| 1 |
+
{"model_name":"ChartGalaxy++ Image2SceneGraph","status":"release","checkpoint_step":100000,"architecture":"Qwen3_5ForConditionalGeneration","backbone_family":"Qwen3.5-4B","weights_format":"safetensors","model_files":[{"path":"chat_template.jinja","sha256":"a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715","size":7756},{"path":"config.json","sha256":"881208c0f3f48d8fe920b4caf1bac23b4e2d351557bf2c1314b2040ceb4c6bba","size":2852},{"path":"model-00001-of-00002.safetensors","sha256":"39c4d658c5e83be2ed19b25e80818f231eff2c147a88398fb6df239a9916326e","size":4973340248},{"path":"model-00002-of-00002.safetensors","sha256":"43a5930fa5f7af81ad51894836f71f124a4d4ef1a21cc3e8c6c75beb8cf14865","size":4346480856},{"path":"model.safetensors.index.json","sha256":"ad380c647e34236cf9f313ff673781e2fca1aaff85af6117727762224f019993","size":67340},{"path":"preprocessor_config.json","sha256":"27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516","size":390},{"path":"processor_config.json","sha256":"d89ef49ce9cd37fbf510158e13c1ef063d9286411c1ec9049932dbe0487143b1","size":1191},{"path":"tokenizer.json","sha256":"06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523","size":19989325},{"path":"tokenizer_config.json","sha256":"792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b","size":1165}],"model_file_bytes":9339891123,"checkpoint_export_file_hashes_match":true,"structure":{"tensor_count":738,"tensor_dtypes":{"BF16":738},"stored_scalar_count":{"BF16":4659865088},"tensor_bytes":9319730176,"weight_map_complete":true,"tensor_offsets_contiguous_and_in_bounds":true,"header_metadata":null},"model_license_status":"owner_selected_cc_by_nc_4_0","local_gpu_inference_smoke":"passed_standalone_example","paper_results_percent":{"element_f1":89.4,"hierarchy_f1":85.1,"spatial_f1":88.0},"paper_test_records":1000,"license":"cc-by-nc-4.0","local_gpu_inference_validation":"smoke/validation.json","spatial_scoring_revision":"geometric-boolean-proximity-v1"}
|
smoke/validation.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"status": "passed",
|
| 3 |
"date_utc": "2026-09-24T12:34:21.923430+00:00",
|
| 4 |
-
"scope": "One standalone image-to-
|
| 5 |
"runtime": {
|
| 6 |
"python": "3.11.14",
|
| 7 |
"vllm": "0.20.2",
|
|
|
|
| 1 |
{
|
| 2 |
"status": "passed",
|
| 3 |
"date_utc": "2026-09-24T12:34:21.923430+00:00",
|
| 4 |
+
"scope": "One standalone image-to-scene-graph inference; no benchmark accuracy evaluation.",
|
| 5 |
"runtime": {
|
| 6 |
"python": "3.11.14",
|
| 7 |
"vllm": "0.20.2",
|