fang-haoshu / halpe-fullbody Goto Github PK

View Code? Open in Web Editor NEW

315.0 315.0 32.0 9.23 MB

Halpe: full body human pose estimation and human-object interaction detection dataset

Python 3.73% Makefile 0.01% Jupyter Notebook 94.43% C++ 0.69% C 0.55% Cython 0.60%

halpe-fullbody's People

Contributors

Stargazers

Watchers

halpe-fullbody's Issues

FileNotFoundError: [Errno 2] No such file or directory: 'halpe_train_v1.json'

Hello there,
I'm trying to execute the vis.py, but halpe_train_v1.json is missing. Where do I get this file from?
Thanks in advanse!
Adriana

Which annotation tool is used to annotate Haple-FullBody dataset?

Thanks for your contribution. Which annotation tool did you use to annotate Haple-FullBody dataset?

There are 41712 image annotations in halpe_train_v1.json, but in HICO-DET dataset, it contains 47,776 images (38,118 in train set and 9,658 in test set)

Thus, which image in HICO-DET dataset should be used corresponding to halpe_train_v1.json?

are the images in this dataset from the COCO dataset

Hello,
Thanks you for creating a wonderful dataset.

I would like to ask if the images from the Halpe-Fullbody dataset are the ones from the COCO2017 dataset?
What is the difference between the Coco Full Body keypoint dataset and Halpe-FullBody dataset? Do they use the same COCO images or not?
Thanks

Only single person annotated per each image

Thank you for sharing the detailed annotations. I downloaded the HICO-DET dataset and the annotations as instructed in the README.md in the repository root. When inspecting the data, I found that only at most one person for each image was annotated. The question is: was it on purpose to annotate only one (not always the main) person or has there been a bug when compiling the annotations jsons? Please see the attached image.
.
I believe this is a duplicate of #3, but I decided to re-open this as a new issue because the original didn't provide enough information.

Here's a minimal script to reproduce, edited from vis.py that came along with the repo.

import os
import json
import cv2
import skimage.io as io
import numpy as np 
from tqdm import tqdm


l_pair = [
    (0, 1), (0, 2), (1, 3), (2, 4),  # Head
    (5, 18), (6, 18), (5, 7), (7, 9), (6, 8), (8, 10),# Body
    (17, 18), (18, 19), (19, 11), (19, 12),
    (11, 13), (12, 14), (13, 15), (14, 16),
    (20, 24), (21, 25), (23, 25), (22, 24), (15, 24), (16, 25),# Foot
    (26, 27),(27, 28),(28, 29),(29, 30),(30, 31),(31, 32),(32, 33),(33, 34),(34, 35),(35, 36),(36, 37),(37, 38),#Face
    (38, 39),(39, 40),(40, 41),(41, 42),(43, 44),(44, 45),(45, 46),(46, 47),(48, 49),(49, 50),(50, 51),(51, 52),#Face
    (53, 54),(54, 55),(55, 56),(57, 58),(58, 59),(59, 60),(60, 61),(62, 63),(63, 64),(64, 65),(65, 66),(66, 67),#Face
    (68, 69),(69, 70),(70, 71),(71, 72),(72, 73),(74, 75),(75, 76),(76, 77),(77, 78),(78, 79),(79, 80),(80, 81),#Face
    (81, 82),(82, 83),(83, 84),(84, 85),(85, 86),(86, 87),(87, 88),(88, 89),(89, 90),(90, 91),(91, 92),(92, 93),#Face
    (94,95),(95,96),(96,97),(97,98),(94,99),(99,100),(100,101),(101,102),(94,103),(103,104),(104,105),#LeftHand
    (105,106),(94,107),(107,108),(108,109),(109,110),(94,111),(111,112),(112,113),(113,114),#LeftHand
    (115,116),(116,117),(117,118),(118,119),(115,120),(120,121),(121,122),(122,123),(115,124),(124,125),#RightHand
    (125,126),(126,127),(115,128),(128,129),(129,130),(130,131),(115,132),(132,133),(133,134),(134,135)#RightHand
]
p_color = [(0, 255, 255), (0, 191, 255), (0, 255, 102), (0, 77, 255), (0, 255, 0),  # Nose, LEye, REye, LEar, REar
           (77, 255, 255), (77, 255, 204), (77, 204, 255), (191, 255, 77), (77, 191, 255), (191, 255, 77),  # LShoulder, RShoulder, LElbow, RElbow, LWrist, RWrist
           (204, 77, 255), (77, 255, 204), (191, 77, 255), (77, 255, 191), (127, 77, 255), (77, 255, 127),  # LHip, RHip, LKnee, Rknee, LAnkle, RAnkle, Neck
           (77, 255, 255), (0, 255, 255), (77, 204, 255),  # head, neck, shoulder
           (0, 255, 255), (0, 191, 255), (0, 255, 102), (0, 77, 255), (0, 255, 0), (77, 255, 255)] # foot

line_color = [(0, 215, 255), (0, 255, 204), (0, 134, 255), (0, 255, 50),
              (0, 255, 102), (77, 255, 222), (77, 196, 255), (77, 135, 255), (191, 255, 77), (77, 255, 77),
              (77, 191, 255), (204, 77, 255), (77, 222, 255), (255, 156, 127),
              (0, 127, 255), (255, 127, 77), (0, 77, 255), (255, 77, 36), 
              (0, 77, 255), (0, 77, 255), (0, 77, 255), (0, 77, 255), (255, 156, 127), (255, 156, 127)]


bodyanno = json.load(open('halpe_train_v1.json'))
image_folder = 'path/to/train images'

imgs = {}
for img in bodyanno['images']:
    imgs[img['id']] = img


annot = bodyanno['annotations'][5]
if 'keypoints' in annot and type(annot['keypoints']) == list:
    imgname = str(imgs[annot['image_id']]['file_name'])
    img = cv2.imread(os.path.join(image_folder, imgname))
    part_line = {}
    kp = np.array(annot['keypoints'])
    kp_x = kp[0::3]
    kp_y = kp[1::3]
    kp_scores = kp[2::3]

    # Draw keypoints
    for n in range(kp_scores.shape[0]):
        if kp_scores[n] <= 0.6:
            continue
        cor_x, cor_y = int(kp_x[n]), int(kp_y[n])
        part_line[n] = (int(cor_x), int(cor_y))
        if n < len(p_color):
            cv2.circle(img, (int(cor_x), int(cor_y)), 2, p_color[n], -1)
        else:
            cv2.circle(img, (int(cor_x), int(cor_y)), 1, (255,255,255), 2)
    # Draw limbs
    for i, (start_p, end_p) in enumerate(l_pair):
        if start_p in part_line and end_p in part_line:
            start_xy = part_line[start_p]
            end_xy = part_line[end_p]
            if i < len(line_color):
                cv2.line(img, start_xy, end_xy, line_color[i], 2)
            else:
                cv2.line(img, start_xy, end_xy, (255,255,255), 1)

plt.imshow(img); plt.axis('off')

What is the dataset license?

Only single person in one picture>

Thank you for share great project. Question as titles.

The dataset cannot be downloaded

@Fang-Haoshu ， Hello ， The dataset cannot be downloaded。

Visibility Score?

I have noticed that visibility scores are not always in between 0 - 2 range for Halpe-FullBody dataset as in coco dataset, where

0 = can't annotate
1 = can annotate for a hidden joint
2 = can annotate because the joint is visible.

What's your visibility keypoints definition?

为什么halpe里面有多个人，halpe_train_v1.json标注文件才只给了一个人的标注？是我搞错了吗？还是就是一张图片只能一个注释？

Issue with HICO-DET Dataset Link and Image Count

I wanted to bring to your attention an issue I encountered while attempting to access the HICO-DET dataset for my research.

The link provided in the paper for the train set is currently not working and leads to a 404 error. Could you please check and update the link?

2.In my attempt to download the images from the HICO website (https://websites.umich.edu/~ywchao/hico/), I noticed that the number of images in the "train2015" folder is 38,118. This differs from the counts mentioned in your paper (40k or 50k). Could you clarify the correct number of images in the train set?

Upon analyzing the annotation files, I observed some repeated entries. After filtering out duplicates, I found a total of 36,120 annotated images. Could you confirm if this is the accurate count or if there's an issue with the annotation files?

Thank you for your time.

Invalid num_keypoints field value

Firstly thanks for sharing the dataset.

While looking into the data in perticular the annotaions dictionary which is suppose to follow the coco format has
invalid values for the field num_keypoints.

All columns values seems to indicate a constant 3 value. the coco format for keypoints indicate this field should
be the sum of keypoints with visibility > 0 but seems not to be the case:

@Fang-Haoshu @HaoyiZhu

Annotation format for bbox is not uniform in training set

Hello,

I found that most bbox annotations are ordered as [x1, y1, width, height], but some images seem to have the annotation like [x1, x2, y1, y2], like "HICO_train2015_00012054.jpg" below. How can I standardize these annotations?

Thanks for your time!

fang-haoshu / halpe-fullbody Goto Github PK

halpe-fullbody's People

Contributors

Stargazers

Watchers

Forkers

halpe-fullbody's Issues

Recommend Projects

Recommend Topics

Recommend Org