yfan1997 commited on
Commit
2b6f484
1 Parent(s): 0ffa7e1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +39 -10
README.md CHANGED
@@ -1,20 +1,32 @@
1
- **Dataset for Aerial Vision-and-Dialog Navigation (AVDN)**
2
 
3
- This is the AVDN dataset proposed in paper "Aerial Vision-and-Dialog Navigation"
4
 
5
- Paper: https://arxiv.org/pdf/2205.12219
 
6
 
7
- Project webpage: https://sites.google.com/view/aerial-vision-and-dialog/home
8
 
9
- We introduced ANDH and ANDH-Full tasks. The ANDH task takes as input the **sub-trajectory** data and you may find instructions to download them at our project webpage.
10
- The ANDH-Full task needs the **full trajectory** data as input, which is shown in the Huggingface repo.
11
 
12
- The data shown in the dataviewer (csv format) is the **full trajectory** data before splited by dialog turns. Please use the following script to convert the csv data into the json format, so that it can fit our current vision of training and inference code.
 
13
 
 
 
14
 
15
- ```
 
 
 
 
 
 
 
 
16
  import pandas as pd
17
  import json
 
18
  # Process each CSV file split
19
  for split in ['train', 'val_seen', 'val_unseen', 'test_unseen']:
20
  # Load the CSV data
@@ -42,7 +54,24 @@ for split in ['train', 'val_seen', 'val_unseen', 'test_unseen']:
42
  json_data.append(entry)
43
 
44
  # Save the data to a JSON file
45
- json_output_path = f'./{split}_full_data_from_csv_test.json'
46
  with open(json_output_path, 'w') as f:
47
  json.dump(json_data, f, indent=4)
48
- ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Dataset for Aerial Vision-and-Dialog Navigation (AVDN)
2
 
3
+ This repository contains the Aerial Vision-and-Dialog Navigation (AVDN) dataset, proposed in the paper "[Aerial Vision-and-Dialog Navigation](https://arxiv.org/pdf/2205.12219)." The dataset and associated tasks enable training and evaluation of models designed for navigation tasks guided by visual and dialog-based cues.
4
 
5
+ - **Paper**: [Aerial Vision-and-Dialog Navigation](https://arxiv.org/pdf/2205.12219)
6
+ - **Project Webpage**: [Aerial Vision-and-Dialog Navigation Project](https://sites.google.com/view/aerial-vision-and-dialog/home)
7
 
8
+ ## Tasks
9
 
10
+ We introduce two tasks within the AVDN framework:
 
11
 
12
+ ### ANDH task
13
+ The **ANDH task** involves **sub-trajectory** data, where navigation occurs over smaller segments of the overall route. Instructions for downloading the dataset related to this task are available on the project webpage.
14
 
15
+ ### ANDH-Full task
16
+ The **ANDH-Full task** uses the **full trajectory** data, providing a more comprehensive view of the navigation route, from start to destination, __which is shown in this repo__ .
17
 
18
+ ## Dataset format
19
+
20
+ The data in the CSV files provided in this repository represents the **full trajectory** data before it is split by dialog turns. The CSV format includes detailed information about each trajectory, including navigation instructions, GPS data, and path-related information.
21
+
22
+ To use this data for training and inference according to our current vision, you'll need to convert it from CSV format to JSON format. Below is a script that converts the CSV data into the required JSON format.
23
+
24
+ ## CSV to JSON conversion script
25
+
26
+ ```python
27
  import pandas as pd
28
  import json
29
+
30
  # Process each CSV file split
31
  for split in ['train', 'val_seen', 'val_unseen', 'test_unseen']:
32
  # Load the CSV data
 
54
  json_data.append(entry)
55
 
56
  # Save the data to a JSON file
57
+ json_output_path = f'./{split}_full_data_from_csv.json'
58
  with open(json_output_path, 'w') as f:
59
  json.dump(json_data, f, indent=4)
60
+
61
+ ```
62
+
63
+ ## Explanations of the key fields in the dataset
64
+
65
+ Below are explanations of the key fields in the dataset, with all coordinates provided in the format [Latitude, Longitude]:
66
+
67
+ - **`map_name`**: The name of the satellite image file corresponding to the environment in which the navigation occurs.
68
+ - **`route_index`**: The index of the specific trajectory being followed by the agent within the map.
69
+ - **`instructions`**: Step-by-step natural language instructions guiding the agent along the route.
70
+ - **`gps_botm_left`**: GPS coordinates representing the bottom-left corner of the map's bounding box.
71
+ - **`gps_top_right`**: GPS coordinates representing the top-right corner of the map's bounding box.
72
+ - **`lng_ratio`**: The ratio used to scale the map's longitude (horizontal) distance to the corresponding pixel dimensions in the image.
73
+ - **`lat_ratio`**: The ratio used to scale the map's latitude (vertical) distance to the corresponding pixel dimensions in the image.
74
+ - **`angle`**: The initial heading direction of the drone at the start of the route. This is expressed in degrees, with 0° being East, 90° being North, 180° being West, and 270° being South.
75
+ - **`destination`**: The final destination of the trajectory, represented as a coordinate or location within the map.
76
+ - **`attention_list`**: A list of human-annotated areas that were the focus during data collection. Each area is represented by a coordinate (latitude, longitude) and a radius (in meters) defining a circular region of interest.
77
+ - **`gt_path_corners`**: A list of four corner points that represent the boundary of each view area along the route. The corners are listed in the following order: forward left corner, forward right corner, backward right corner, and backward left corner. This provides the ground truth view area at each step of the navigation.