> For the complete documentation index, see [llms.txt](https://spoofsense.gitbook.io/spoofsense-triton-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://spoofsense.gitbook.io/spoofsense-triton-documentation/image-requirements.md).

# Image requirements

The API expects the input image to have certain requirements in order to accurately perform facial liveness checks.

### Best practices 

* Supported image formats: png, jpg, jpeg.
* Image size should not exceed: 10mb
* Image should be clicked in a well-lit room, with the face shown clearly to the camera. Extreme darkness/brightness on the face might increase the BPCER.
* The Face should be straight and not tilted. 
* Minimum Size of detected face should be 224X224 pixels. (APCER will be affected for very small face bboxes)
* The padding between the face box and the image's borders should be at least 15 pixels. 
* Distance between eyes should be ateast 100 pixels.
* Motion Blur might increase BPCER.
* Any occlusions on parts of face like Glasses, Masks should be removed while performing Liveness Check.
* The detected Face should occupy atleast 20% of the total image area.
* Image should be clicked as a selfie (portrait mode only). 
* No image compression is highly recommended.

### Compressing Images for high performance - python

Use the following snippet to reduce the size of the images before sending it to the spoofsense-api.

```python
from PIL import Image

def compress_image(image_path, output_path, quality):
    # Open the image file.
    with Image.open(image_path) as img:
        # Save the image again with reduced quality.
        img.save(output_path, "JPEG", quality=quality)

# Path of the image to be compressed.
image_path = "path/to/your/image.jpg"

# Path where the compressed image will be saved.
output_path = "path/to/your/compressed/image.jpg"

# Quality parameter. Lower value will reduce the size more, but may decrease the quality.
# In many cases, values around 80 to 90 still provide good quality images.
quality = 90

compress_image(image_path, output_path, quality)

```

### <br>