PointNet:
PointNet trained on the dataset ShapeNet (airplane), by using the OpenMLCore framework. The based Pytorch Code is available in model.py
📌 Results
Visualisations:
There are some results after the training on Augmented_Dubai-Satellite_Image_segmentation dataset.
Airplane subdataset Multiclass:
| Result 1 | Result 2 |
|---|---|
 |
 |
📊 Metrics & Performances
Performance on ShapeNet (Airplane) dataset:
There are the performances of the model on the Augmented_Dubai-Satellite_Image_segmentation dataset. Please note that, on Dubai Satellite dataset, we have a class imbalance for water(class0), road(class2), Building (class3) for this dataset...
| Model / Configuration | Accuracy |
|---|---|
| Model V1 | 0.9121 |
⚙️ Model training:
I used my personal Framework ( a wrapper of PyTorch) to train the model.
Install the framework:
git clone https://github.com/sebastien-doyez2812/OpenML-core.git
cd OpenML-core