Help request

[Nizar150]

I had a question and was hoping you could help with it .

I work on a project which include a part where i have to count/detect riyads( which are a typr of traditional housing in Morocco) from aerial images and i was using sam geospatial and i wanted to ask if you had an idea if it s possible to automate the segmentation of these houses ( there is a methode where you select manualy the masks but i wanted to know if there was way to do it automaticaly like choosing some sample data but then let it do the rest) and the final result would be two mask regulars house/ building and those riyad .

I hope i can get an anwser from you
thank you .

[trevorhlynn]

Hello! We have some resources that could be helpful:

• GitHub - autodistill/autodistill: Images to inference with no labeling (use foundation models to train supervised models)
• How to Use the Segment Anything Model (SAM)
• Grounding DINO : SOTA Zero-Shot Object Detection
• Train a Model in Roboflow - Roboflow Docs
• How to Train YOLOv8 Instance Segmentation on Custom Data

These are tools to automate applying segmentation masks, label an initial dataset to train using Roboflow Train or the final link is a way to train your own custom YOLOv8 model. Depending on the level of effort you’d like to pursue, one of these options should work.