Image augmentation preview
Upload an image and instantly see the 10 most-used augmentation transforms side-by-side: horizontal/vertical flip, rotation, brightness, contrast, grayscale, blur, hue rotation, and Gaussian noise. Everything runs in your browser — your image never leaves your machine.
About these transforms
This preview uses native Canvas API filters and pixel-level maths — the same logic, structurally, as Albumentations applies at training time (just simpler).
- Flip / Rotate — geometric transforms; in production these also transform your bounding boxes and masks.
- Brightness / Contrast / Hue — colour-space jitter for lighting and white-balance robustness.
- Grayscale — probability-gated random grayscale; helps when your model shouldn't rely on colour cues.
- Blur — simulates out-of-focus shots, fast motion, sensor blur. Stronger than typical training defaults.
- Gaussian noise — additive pixel noise simulating low-light sensor noise. Use modestly; over-noised images train brittle models.
The same pipeline, at dataset scale
This page previews one image at a time. The mSightFlow augmentation API applies your declared pipeline across the whole dataset, with bbox-aware transforms so labels move with the image — and exports the result as COCO or YOLO.
# Pipeline as JSON — Albumentations under the hood
pipeline = [
{"type": "HorizontalFlip", "p": 0.5},
{"type": "Rotate", "limit": 15, "p": 0.7},
{"type": "RandomBrightnessContrast", "brightness_limit": 0.2, "p": 0.6},
{"type": "GaussNoise", "var_limit": [10, 50], "p": 0.3},
]
requests.post(
"https://api.msightflow.ai/v1/projects/PROJECT_ID/export-augmented",
headers={"Authorization": f"Bearer {os.environ['MSF_API_KEY']}"},
json={
"pipeline": pipeline,
"augmentations_per_image": 4,
"format": "yolo",
"split": {"train": 0.8, "val": 0.1, "test": 0.1},
},
)Sibling tools
Augment your whole dataset — bbox-aware.
50 free exports / month. Bbox-aware albumentations, train/val/test split, webhook on completion.