1. Depth Prediction Dataset Format

1.1 Raw Image Sets

    The raw image sets contain RGB images in PNG format, each of which is in resolution of 500 (height) * 1000 (width). The raw image set "ParallelEye_rgb_train.rar" and corresponding depth prediction image set "ParallelEye_depth_train.rar" can be downloaded from the homepage. In depth prediction task, the depth prediction images are provided only for the images in folders "04", "05" and "06".

1.2 Ground Truth Sets

    For the raw image sets, the ground truth of each training image is provided also as 24 bit(8 bit/channel) RGB image in PNG format. Since the intensity values of three channels are equal, the ground truth depth information can be obtained from any channel.
    The depth for a pixel can be computed in meters by:

    Figure 1 shows an example of ground truth depth for an image.

Figure 1. Example of depth data. Left: raw image. Right: depth ground truth.

2. Depth Prediction Task

    For each test image, you should predict the depth of all pixels.
    The output from your system should be an 8 bit image in PNG format.

3. Evaluation

3.1 Data Supplied

    The test sets ("segmentation&depth_testing_set.rar" in homepage) contain 1022 RGB images in the same PNG format with training images.

3.2 Submission of Results

    The test results should be submitted as collections of single-channel 8 bit PNG image files. One PNG image file should be generated for each test image and named in the same way.

3.3 Evaluation method

    Each depth prediction competition will be evaluated with the root mean squared error (RMSE) of the inverse depth maps . For a predicted depth map y and ground truth y*, each with |T| pixels indexed by i,

    Participants are expected to submit a single set of test results for your team.