Pennfudan

class olympus.datasets.pennfudan.CocoDetection(img_folder, ann_file, transforms)[source]

Bases: torchvision.datasets.coco.CocoDetection

Methods

extra_repr  
register_datapipe_as_function  
register_function  
class olympus.datasets.pennfudan.PennFudanDataset(data_path)[source]

Bases: olympus.datasets.dataset.AllDataset

This is an image database containing images that are used for pedestrian detection in the experiments reported in [1]. The images are taken from scenes around campus and urban street. The objects we are interested in these images are pedestrians. Each image will have at least one pedestrian in it.

All labeled pedestrians are straight up. More on official website.

References

[1]Liming Wang, Jianbo Shi, Gang Song, I-fan Shen. “Object Detection Combining Recognition and Segmentation”. ACCV 2007
Attributes:
input_shape: (3, H, W) with H ∈ [311, 581], W ∈ [253, 1017] and H * W ∈ [81719, 451548]

The heights of labeled pedestrians in this database fall into [180,390] pixels. They are all have unique shapes

target_shape: DictionaryKeys(‘boxes’, ‘labels’, ‘masks’, ‘image_id’, ‘area’, ‘iscrowd’)

boxes: Tensor[P, 4] where P equals the number of pedestrian and 4 is the bounding box labels: Tensor[P], always 1 masks: Tensor[P, H, W], 1 when the pixel belongs to a pedestrian, 0 if not image_id: Tensor[1], image id inside the dataset iscrowd: Tensor[P], if the image has a crowd of people (always false)

train_size: 136

Size of the train dataset 96 images are taken from around University of Pennsylvania 74 are taken from around Fudan University.

valid_size: 16

Size of the validation dataset

test_size: 16

Size of the test dataset

Methods

categories() Dataset tags so we can filter what we want depending on the task
transforms()
collate_fn  
register_datapipe_as_function  
register_function  
static categories()[source]

Dataset tags so we can filter what we want depending on the task

collate_fn()
num_classes
olympus.datasets.pennfudan.penn_collate_fn(batch)[source]