tinydataml / tiny3d Goto Github PK

• support tiny3d.lidar_preprocess(lidar_data, method_name)

support

• tiny3d.lidar_outlying_detection(dataset, od_method_name)
• tiny3d.lidar_out_of_distribution_detection(dataset, ood_method_name)

support tiny3d.pseudo_label(test_data_path, pretrained_model, output_test_data_path)

• semi-supervised learning

• support tiny3d.deploy(centerpoint_model, 'onnx')
• support tiny3d.deploy(pointpillars_model, 'onnx')
• support colab demo

• support tiny3d.train(dataset, model, trainer)
• support tiny3d.test(dataset, model)

• support tiny3d.compress(model, compression_method_name)
• support tiny3d.compress(model, compression_method_name)
• support tiny3d.compress_predicted_latency(model, compression_method_name)
• support tiny3d.compress_predicted_size(model, compression_method_name)
• support tiny3d.compress_predicted_flops(model, compression_method_name)
• support tiny3d.compress_predicted_auuracy(model, compression_method_name)

support

• tiny3d.lidar_evaluation(lidar_data, smooth_method_name)

• support tiny3d.long_tail(lidar_data, method_name)

support

• tiny3d.lidar_evaluation(lidar_data, evaluation_method_name)

support tiny3d.compress(dataset, augemtatino_method_name)

• global augmentation methods
• local augementation methods
• frustom augementation methods
• pyramid augmentation methods
• other novel augmentation methods

• https://github.com/pyun-ram/incdet3

support

• tiny3d.domain_adaptation(domain1_data_path, domain2_data_path, model)

support tiny3d.model_ensemble(model_list, ensemble_method_name)

• Non-maximum Suppression (NMS)
• Soft-NMS [1]
• Non-maximum weighted (NMW) [2]
• Weighted boxes fusion (WBF) [3] - new method which gives better results comparing to others

support tiny3d.lidar_simulator(data_path, expanding_method_name)

• Rainy Day
• Snow Day
• Foggy Day

support tiny3d.lidar_torch_dataloader(dataset_path)

• support Front-end interactive tiny3d

• support tiny3d.bad_case_visualization(dataset, model, topk, visulization_method_name)

• support tiny3d.lidar_load(data_path, data_type, load_method_name)

support tiny3d.lidar_cleaning(data_path, cleaning_method_name)

• denoising method

support

• tiny3d.lidar_robust(dataset, robusten_method_name)

• code align

• support tiny3d colab demo

• support tiny3d.deploy(pointpillars_model, 'onnx')
• support colab demo

• support tiny3d.slim_data(dataset, model, active_learning_method_name)

• support tiny3d.deploy(pvrcnn_model, 'onnx')
• support colab demo

support tiny3d.lidar_quality(data_path, quality_metric_name)