HaGRID – HAnd Gesture Recognition Image Dataset

HaGRID – HAnd Gesture Recognition Image Dataset


HaGRID - HAnd Gesture Recognition Image Dataset


HaGRID - HAnd Gesture Recognition Image Dataset

Use Case

Gesture Recognition


The HaGRiD (Hand Gesture Recognition Image Dataset) is a collection of images and data designed for the development and evaluation of hand gesture recognition systems. Hand gesture recognition is a technology used in human-computer interaction, sign language recognition, and various other applications.


HaGRID – HAnd Gesture Recognition Image Dataset

We introduce a large image dataset HaGRID (HAnd Gesture Recognition Image Dataset) for hand gesture recognition (HGR) systems. You can use it for image classification or image detection tasks. Proposed dataset allows to build HGR systems, which can be used in video conferencing services (Zoom, Skype, Discord, Jazz etc.), home automation systems, the automotive sector, etc.

HaGRID size is 716GB and dataset contains 552,992 FullHD (1920 × 1080) RGB images divided into 18 classes of gestures. Also, some images have no_gesture class if there is a second free hand in the frame. This extra class contains 123,589 samples. The data were split into training 92%, and testing 8% sets by subject user-id, with 509,323 images for train and 43,669 images for test.

HaGRID – HAnd Gesture Recognition Image Dataset

 The subjects are people from 18 to 65 years old. The dataset was collected mainly indoors with considerable variation in lighting, including artificial and natural light. 

The annotations consist of bounding boxes of hands with gesture labels in COCO format [top left X position, top left Y position, width, height]. Also, annotations have 21 landmarks in format [x,y] relative image coordinates, markups of leading hands (left of right for gesture hand) and leading_conf as confidence for leading_hand annotation. We provide user_id field that will allow you to split the train / val dataset yourself.
"0534147c-4548-4ab4-9a8c-f297b43e8ffb": {
  "bboxes": [
    [0.38038597, 0.74085361, 0.08349486, 0.09142549],
    [0.67322755, 0.37933984, 0.06350809, 0.09187757]
        [0.39917091, 0.74502739],
        [0.42500172, 0.74984396],
        [0.70590734, 0.46012364],
        [0.69208878, 0.45407018],
  "labels": [
  "leading_hand": "left",
  "leading_conf": 1.0,
  "user_id": "bb138d5db200f29385f..."


Download and unzip our train dataset split into 18 archives by gestures. Get started with the links below


GestureSize    Gesture                      Size
call        39.1 GBpeace                        38.6 GB
dislike   38.7 GBpeace_inverted     38.6 GB
fist         38.0 GB rock                          38.9 GB
four       40.5 GB stop                          38.3 GB
like        38.3 GB stop_inverted        40.2 GB
mute    39.5 GB three                         39.4 GB
ok          39.0 GB three2                      38.5 GB
one       39.9 GB two_up                   41.2 GB
palm    39.3 GB two_up_inverted 39.2 GB
train_val annotations: ann_train_val


Test Archives       Size
images test         60.4 GB
annotations ann_test 27.3 MB

It has 100 items per gesture.

  Archives              Size
images subsample                2.5 GB
annotations ann_subsample 1.2 МB


We provide some pre-trained classifiers and detectors as the baseline solutions. You can download them via github repo.



Globose Technology Solutions excels in hand gesture recognition, utilizing the HaGRID dataset for precise identification. Our advanced algorithms contribute to innovative applications in human-computer interaction and virtual reality, reflecting our commitment to cutting-edge gesture recognition technology.

Contact Us

Please enable JavaScript in your browser to complete this form.

Quality Data Creation


Guaranteed TAT


ISO 9001:2015, ISO/IEC 27001:2013 Certified


HIPAA Compliance


GDPR Compliance


Compliance and Security

Let's Discuss your Data collection Requirement With Us

To get a detailed estimation of requirements please reach us.

Scroll to Top

Please provide your details to download the Dataset.