Skip to main content

An educational module for experimenting with single-instance and multi-instance object detection and for generating region proposals with graph-based algorithms

Project description

Consult the module API page at

https://engineering.purdue.edu/kak/distRPG/RegionProposalGenerator-2.0.2.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
    RegionProposalGenerator class and invoke the methods shown below on
    this instance:

    rpg = RegionProposalGenerator(
                      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 = RegionProposalGenerator.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:

    To generate region proposals, you would need to construct an instance
    of the RegionProposalGenerator class and invoke the methods shown below
    on this instance:

    rpg = RegionProposalGenerator(
                   ###  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" )

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page