Why VisionGuard?

An imperceptible perturbation can cause your vision model to fail. In healthcare, agriculture, industry, or surveillance, the consequences can be critical. VisionGuard tests your models against the full spectrum of known adversarial attacks, locally, without ever accessing your weights.

Supported model types

Classification

Single-label and multi-label image classification

ResNet · VGG · MobileNet · EfficientNet · ViT · ConvNeXt · DenseNet · Inception

Object Detection

Single-class and multi-class detection with bounding boxes

YOLOv5 · YOLOv8 · YOLOv10 · YOLOv11 · Faster R-CNN · RetinaNet · DETR · SSD

Semantic Segmentation

Pixel-wise classification by class

U-Net · DeepLabv3 · SegFormer · Mask2Former · PSPNet · FCN

Instance Segmentation

Per-instance segmentation

Mask R-CNN · YOLACT · SOLO · SAM · YOLOv8-seg · DETR-seg

Pose Estimation

Human or object keypoint detection

OpenPose · HRNet · MediaPipe · YOLOv8-pose · MoveNet

Depth Estimation

Monocular and stereo depth

MiDaS · DPT · ZoeDepth · Depth Anything · MonoDepth

Face Recognition & Biometrics

Identification, verification, spoofing detection

FaceNet · ArcFace · MTCNN · InsightFace

Medical & Specialized Imaging

Models trained on medical, satellite, agricultural, industrial imagery

U-Net 3D · nnU-Net · Custom models

Super-Resolution & Restoration

Image enhancement, denoising, deblurring

ESRGAN · SwinIR · Real-ESRGAN · DnCNN

Generative Models

GANs, autoencoders, diffusion models (deepfake detection)

StyleGAN · Stable Diffusion · VAE · Deepfake detectors

OCR & Document Reading

Text recognition, document analysis, codes

EasyOCR · PaddleOCR · TrOCR · LayoutLM

Your model not listed?

If your architecture or use case isn't listed, contact us. We develop test modules on demand.

Contact us →

Attack categories tested

Perturbation Attacks

Imperceptible modifications added to input to fool the model. Tests at varying noise levels.

Optimization Attacks

Iterative attacks that find the minimal perturbation needed to break the model.

Detection Evasion

Attacks specific to object detectors: object vanishing, false positives, NMS bypass.

Gradient Masking Diagnostic

False robustness detection: your model appears robust but only masks its gradients. Methodology from our ongoing research (Springer paper).

Adversarial Patches

Tests with printable visible patches to evaluate robustness in real-world conditions.

Physical Transformations

Robustness against rotations, lighting changes, partial occlusions, Gaussian noise.

Get started in 3 commands

# Installpip install rednblue

# Authenticateset RNB_TOKEN=your_token

# Run a test (Classifier)
rnb preview --model resnet50.pth --input images/ --submit

# For YOLOrnb preview --model-type yolo --model best.pt --input images/ --submit

Ready to test your vision models?