The input of this model is expected to be a chart image. You can use the Nemotron Page Element v3 to detect and crop such images.
Description
The Nemotron Graphic Elements v1 model is a specialized object detection system designed to identify and extract key elements from charts and graphs. Based on YOLOX, an anchor-free version of YOLO (You Only Look Once), this model combines a simpler architecture with enhanced performance. While the underlying technology builds upon work from Megvii Technology, we developed our own base model through complete retraining rather than using pre-trained weights.
The model excels at detecting and localizing various graphic elements within chart images, including titles, axis labels, legends, and data point annotations. This capability makes it particularly valuable for document understanding tasks and automated data extraction from visual content.
This model is ready for commercial/non-commercial use.
We are excited to announce the open sourcing of this commercial model. For users interested in deploying this model in production environments, it is also available via the model API in NVIDIA Inference Microservices (NIM) at nemotron-graphic-elements-v1.
The Nemotron Graphic Elements v1 is designed for automating extraction of graphic elements of charts in enterprise documents. Key applications include:
Enterprise document extraction, embedding and indexing
Augmenting Retrieval Augmented Generation (RAG) workflows with multimodal retrieval
Architecture Type: YOLOX Network Architecture: DarkNet53 Backbone + FPN Decoupled head (one 1x1 convolution + 2 parallel 3x3 convolutions (one for the classification and one for the bounding box prediction). YOLOX is a single-stage object detector that improves on Yolo-v3. This model was developed based on the Yolo architecture Number of model parameters: 5.4e7
Input
Input Type(s): Image Input Format(s): Red, Green, Blue (RGB) Input Parameters: Two-Dimensional (2D) Other Properties Related to Input: Image size resized to (1024, 1024)
Output
Output Type(s): Array Output Format: A dictionary of dictionaries containing np.ndarray objects. The outer dictionary has entries for each sample (page), and the inner dictionary contains a list of dictionaries, each with a bounding box (np.ndarray), class label, and confidence score for that page. Output Parameters: One-Dimensional (1D) Other Properties Related to Output: The output contains bounding boxes, detection confidence scores, and object classes (chart title, x/y axis titles and labels, legend title and labels, marker labels, value labels and other texts). The thresholds used for non-maximum suppression are conf_thresh=0.01 and iou_thresh=0.25.
Output Classes:
Chart title
Title or caption associated to the chart
x-axis title
Title associated to the x axis
y-axis title
Title associated to the y axis
x-axis label(s)
Labels associated to the x axis
y-axis label(s)
Labels associated to the y axis
Legend title
Title of the legend
Legend label(s)
Labels associated to the legend
Marker label(s)
Labels associated to markers
Value label(s)
Labels associated to values
Other
Miscellaneous other text components
Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.
Usage
The model requires torch, and the custom code available in this repository.
Optional:
This can be installed as a package using pip
cd nemotron-graphic-elements-v3
pip install -e .
Run the model using the following code:
import torch
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
from nemotron_graphic_elements_v1.model import define_model
from nemotron_graphic_elements_v1.utils import plot_sample, postprocess_preds_graphic_element, reformat_for_plotting
# Load image
path = "./example.png"
img = Image.open(path).convert("RGB")
img = np.array(img)
# Load model
model = define_model("graphic_element_v1")
# Inference
with torch.inference_mode():
x = model.preprocess(img)
preds = model(x, img.shape)[0]
print(preds)
# Post-processing
boxes, labels, scores = postprocess_preds_graphic_element(preds, model.threshold, model.labels)
# Plot
boxes_plot, confs = reformat_for_plotting(boxes, labels, scores, img.shape, model.num_classes)
plt.figure(figsize=(15, 10))
plot_sample(img, boxes_plot, confs, labels=model.labels)
plt.show()
Note that this repository only provides minimal code to infer the model.
If you wish to do additional training, refer to the original repo.
Advanced post-processing
Additional post-processing might be required to use the model as part of a data extraction pipeline.
We provide examples in the notebook Demo.ipynb.
Model Version(s):
nemotron-graphic-elements-v1
Training and Evaluation Datasets:
Training Dataset
Data Modality: Image Image Training Data Size: Less than a Million Images Data collection method by dataset: Automated Labeling method by dataset: Hybrid: Automated, Human Pretraining (by NVIDIA): 118,287 images of the COCO train2017 dataset Finetuning (by NVIDIA): 5,614 images from the PubMed Central (PMC) Chart Dataset. 9,091 images from the DeepRule Dataset with annotations obtained using the CACHED model Number of bounding boxes per class:
Label
Images
Boxes
chart_title
9,487
18,754
x_title
5,995
9,152
y_title
8,487
12,893
xlabel
13,227
217,820
ylabel
12,983
172,431
legend_title
168
209
legend_label
9,812
59,044
mark_label
660
2,887
value_label
3,573
65,847
other
3,717
29,565
Total
14,143
588,602
Evaluation Dataset
Results were evaluated using the PMC Chart dataset. The Mean Average Precision (mAP) was used as the evaluation metric to measure the model's ability to correctly identify and localize objects across different confidence thresholds.
Number of bounding boxes and images per class:
Label
Images
Boxes
chart_title
38
38
x_title
404
437
y_title
502
505
xlabel
553
4,091
ylabel
534
3,944
legend_title
17
19
legend_label
318
1,077
mark_label
42
219
value_label
52
726
other
113
464
Total
560
11,520
Data collection method by dataset: Hybrid: Automated, Human Labeling method by dataset: Hybrid: Automated, Human Properties: The validation dataset is the same as the PMC Chart dataset.
Per-class Performance Metrics:
Class
AP (%)
AR (%)
chart_title
82.38
93.16
x_title
88.77
92.31
y_title
89.48
92.32
xlabel
85.04
88.93
ylabel
86.22
89.40
other
55.14
79.48
legend_label
84.09
88.07
legend_title
60.61
68.42
mark_label
49.31
73.61
value_label
62.66
68.32
Ethical Considerations
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
For more detailed information on ethical considerations for this model, please see the Explainability, Bias, Safety & Security, and Privacy sections below.
Please report security vulnerabilities or NVIDIA AI Concerns here.
Bias
Field
Response
Participation considerations from adversely impacted groups protected classes in model design and testing
None
Measures taken to mitigate against unwanted bias
None
Explainability
Field
Response
Intended Application & Domain:
Object Detection
Model Type:
YOLOX-architecture for detection of graphic elements within images of charts.
Intended User:
Enterprise developers, data scientists, and other technical users who need to extract textual elements from charts and graphs.
Output:
After post-processing, the output is three numpy array that contains the detections: boxes [N x 4] (format is normalized (x_min, y_min, x_max, y_max)), associated classes: labels [N] and confidence scores: scores [N].
Describe how the model works:
Finds and identifies objects in images by first dividing the image into a grid. For each section of the grid, the model uses a series of neural networks to extract visual features and simultaneously predict what objects are present (in this case "chart title" or "axis label" etc.) and exactly where they are located in that section, all in a single pass through the image.
Name the adversely impacted groups this has been tested to deliver comparable outcomes regardless of:
Not Applicable
Technical Limitations & Mitigation:
The model may not generalize to unknown chart types/formats. Further fine-tuning might be required for such images.
Verified to have met prescribed NVIDIA quality standards:
Yes
Performance Metrics:
Mean Average Precision, detectionr recall and visual inspection
Potential Known Risks:
This model may not always detect all elements in a document.
The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to.
