PG-SceneParse150-M - a computer-vision model on Pictograph

Open in task arch mAP license

View on Pictograph · Pictograph Research · Instance Segmentation · RF-DETR Seg Medium · FP32 · BSD 3-Clause License

Overview

This is an instance segmentation model built on RF-DETR Seg Medium, trained on the ADE-20K dataset. It expects 432x432 input and runs in FP32 precision. The ONNX weights are in this repo; run them with the Pictograph SDK (below) or open the model on Pictograph to test it in-browser, call the hosted API, or deploy it as an always-on endpoint.

Performance

Headline: 26.4% mAP.

Metric Value
mAP 26.4%
mAP@50 41.4%

Classes

Class index matches the model's output order.

# Class
0 wall
1 building
2 sky
3 floor
4 tree
5 ceiling
6 road
7 bed
8 windowpane
9 grass
10 cabinet
11 sidewalk
12 person
13 earth
14 door
15 table
16 mountain
17 plant
18 curtain
19 chair
20 car
21 water
22 painting
23 sofa
24 shelf
25 house
26 sea
27 mirror
28 rug
29 field
30 armchair
31 seat
32 fence
33 desk
34 rock
35 wardrobe
36 lamp
37 bathtub
38 railing
39 cushion
40 base
41 box
42 column
43 signboard
44 chest_of_drawers
45 counter
46 sand
47 sink
48 skyscraper
49 fireplace
50 refrigerator
51 grandstand
52 path
53 stairs
54 runway
55 case
56 pool_table
57 pillow
58 screen_door
59 stairway

Showing the first 60 of 150 classes.

Training

Setting Value
Base dataset ADE-20K
Dataset size 22,207 images
Input size 432x432
Epochs 12
Batch size 4
Learning rate 0.0002
Model size medium
Precision FP32
Version 1.0.0

Use this model

The weights ship as ONNX (pictograph-sceneparse150-m.onnx) with a config.json. The Pictograph SDK applies the model's exact pre/post-processing for you:

Pictograph SDK

pip install "pictograph[inference]"
from huggingface_hub import hf_hub_download
from pictograph import load_model, InstanceSegmentationModel

onnx = hf_hub_download("pictograph/pg-sceneparse150-m", "pictograph-sceneparse150-m.onnx")
config = hf_hub_download("pictograph/pg-sceneparse150-m", "config.json")

model: InstanceSegmentationModel = load_model(onnx, config, task="instance_segmentation")
result = model.predict("photo.jpg")   # InstanceSegmentationResult
for p in result.polygons:              # instances whose mask polygonized
    b = p.bounding_box
    ring = p.polygon.paths[0]          # outer ring: [{'x','y'}, ...]
    print(f"{p.name} {p.confidence:.2f} {len(ring)}-pt mask "
          f"box=({b.x:.0f},{b.y:.0f},{b.w:.0f},{b.h:.0f})")

Example output (illustrative; your exact values depend on the image):

wall 0.91 48-pt mask box=(118,86,140,210)
building 0.87 52-pt mask box=(120,250,132,300)
Raw ONNX Runtime (no SDK)
from huggingface_hub import hf_hub_download
import onnxruntime as ort

onnx = hf_hub_download("pictograph/pg-sceneparse150-m", "pictograph-sceneparse150-m.onnx")
sess = ort.InferenceSession(onnx, providers=["CPUExecutionProvider"])
inp = sess.get_inputs()[0]
print(inp.name, inp.shape)   # NCHW float32; input size + normalization in config.json
# outputs = sess.run(None, {inp.name: preprocessed_batch})
# Decoding is architecture-specific; the Pictograph SDK (above) does it for you.

License

Released under BSD 3-Clause License.

Citation

This model uses the RF-DETR Seg Medium architecture:

@inproceedings{robinson2026rfdetr,
  title     = {RF-DETR: Real-Time Detection Transformer},
  author    = {Robinson, Isaac and Robicheaux, Peter and Popov, Matvei and Ramanan, Deva and Peri, Neehar},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2026},
  url       = {https://arxiv.org/abs/2511.09554}
}

Trained and published with Pictograph - annotate, train, and deploy from one API.

Downloads last month
21
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for pictograph/pg-sceneparse150-m

Finetuned
(6)
this model

Dataset used to train pictograph/pg-sceneparse150-m

Paper for pictograph/pg-sceneparse150-m

Evaluation results