Image Datasets

Quickly expand image training data to build better AI applications

Mindful AI owns image datasets with intellectual property rights including surveillance security,
object classification and attributes, OCR, and faces. In addition, we have annotated these data
with an accuracy rate higher than 95%, such as bounding boxes, semantic segmentation,
key points, classification and attributes, etc.

Image Datasets

27,000 images of 1-pixel accuracy segmentation annotation for body&hair

Image amount: 27,000

Source: web crawling

Annotations: segmentation on body, hair, bag, hat, hair accessories, etc.

Marginal error: within ±1 pixel

Accuracy rate: >97%

Image Datasets

2,000 White People Expression Images

25 images per person, total of

50,000 images

even distribution of male and female

the image width is more than 3024p

Image Datasets

2,000 Black People Expression Images

36 images per person, total of 72,000 images

even distribution of male and female

the image width is more than 1080p

Image Datasets

600 Filipino_Expression, Pose, Lights

Front_normal expression_normal light , Amazed , Angry , Disgusted ,

Happy , Sad , Scared , Indoor_dim light , Indoor_strong light , Outdoor_dim light , Outdoor_strong light , Indoor_down , Indoor_Yaw_left , Indoor_Roll_Left , Indoor_LRight , Indoor_Normal , Indoor_Right , Indoor_Teeth , Indoor_Up , Outdoor_down , Outdoor_left , Outdoor_Yaw_left , Outdoor_Roll_Left , Outdoor_Normal , Outdoor_Right , Outdoor_Teeth , Outdoor_Up , Indoor_closing eye , Outdoor_closing eye , Indoor_with glasses type1 , Indoor_with glasses type2 , Indoor_without glasses , Outdoor_with glasses type1 , Outdoor_with glasses type2 , Outdoor_without glasses

Image Datasets

114,000 images of face and upper body

Image amount: 114K

Source: web crawling

Race: mostly Asian

Capturing environment: both indoor and outdoor

Photo features: with diversities of bag, accessory, sunglasses, clothes/shoes of various styles.

Pixel count: more than 600x800p

All original images, without any annotations

Image Datasets

1,600people_multiple face poses and lights (without annotations)

Image amount: 1,600 people, 120-140 images per each person, total of 200K images

Source: capturing with legal authorization letter signed

Pixel count: above 600×800

Capturing device: smart phones like iPhone, Samsung, Xiaomi, etc. Capturing environment: indoor, outdoor

Light conditions: strong, normal, low light

Poses: opening eyes, closed eyes, head up, head down, left face, right face

Image Datasets

43,131 images from surveillance cam with bounding box annotations

Image amount: 43,131

Capturing device: 1080P surveillance camera

Cam set-up locations: 222 outdoor crossroads in China urban areas

Time: 146 locations captured in daytime and 76 locations captured in night

Annotation contents: bounding box, target name, truncation, hard to distinguish

Output format: xml file (PASCAL VOC)

Annotated targets: person, bus, car, bicycle, motorcycle, truck

Image Datasets

4,474 images of sky

Image amount: 4,474

Capturing device: 1080P surveillance camera

Pixel count: the short side is more than 1080p

Capturing device: smart phones like iPhone, Samsung, Xiaomi, etc.

There are sky, trees, buildings, streetlight, persons, wires in the images.

Image Datasets

1M bbox of OCR_ Thai, Vietnamese, Japanese, Korean

Bounding box amount: 1M

Image types: slogans, business cards, menus, maps, store names, bills, billboards, packaging, handwriting, etc.

Pixel count: above 1280×720

Annotations: bounding box, text transcription

Embrace Your Digital Transformation

Start your journey today