YOLOLogic 2.1.5

Consult the module API page at

https://engineering.purdue.edu/kak/distYOLO/YOLOLogic-2.1.5.html

for all information related to this module, including information related to the latest changes to the code. The page at the URL shown above lists all of the module functionality you can invoke in your own code.

Single-Instance and Multi-Instance Object Detection:

    Say you wish to experiment with YOLO-like logic for multi-instance
    object detection, you would need to construct an instance of the
    YOLOLogic class and invoke the methods shown below on
    this instance:

    rpg = YOLOLogic(
                      dataroot = "./data/",
                      image_size = [128,128],
                      yolo_interval = 20,
                      path_saved_yolo_model = "./saved_yolo_model",
                      momentum = 0.9,
                      learning_rate = 1e-6,
                      epochs = 40,
                      batch_size = 4,
                      classes = ('Dr_Eval','house','watertower'),
                      use_gpu = True,
                  )
    yolo = YOLOLogic.YoloLikeDetector( rpg = rpg )
    yolo.set_dataloaders(train=True)
    yolo.set_dataloaders(test=True)
    model = yolo.NetForYolo(skip_connections=True, depth=8)
    model = yolo.run_code_for_training_multi_instance_detection(model, display_images=False)
    yolo.run_code_for_training_multi_instance_detection(model, display_images = True)


Graph-Based Algorithms for Region Proposals:

    For experimenting with the graph-based methods for image segmentation,
    you would need to construct an instance of the YOLOLogic class and invoke
    the methods shown below on this instance:

    rpg = YOLOLogic(
                   ###  The first 6 options affect only the graph-based part of the algo
                   sigma = 1.0,
                   max_iterations = 40,
                   kay = 0.05,
                   image_normalization_required = True,
                   image_size_reduction_factor = 4,
                   min_size_for_graph_based_blobs = 4,
                   ###  The next 4 options affect only the Selective Search part of the algo
                   color_homogeneity_thresh = [20,20,20],
                   gray_var_thresh = 16000,
                   texture_homogeneity_thresh = 120,
                   max_num_blobs_expected = 8,
          )

    image_name = "images/mondrian.jpg"
    segmented_graph,color_map = rpg.graph_based_segmentation(image_name)
    rpg.visualize_segmentation_in_pseudocolor(segmented_graph[0], color_map, "graph_based" )
    merged_blobs, color_map = rpg.selective_search_for_region_proposals( segmented_graph, image_name )
    rpg.visualize_segmentation_with_mean_gray(merged_blobs, "ss_based_segmentation_in_bw" )