The CASC IFW Database

Overview

Welcome to the Internal Feeding Worm (IFW) Database of the Comprehensive Automation for Specialty Crops (CASC) research project. This project is funded by the USDA Specialty Crop Research Initiative (SCRI) program under award no. 2008-51180-04876..

Project members at Purdue University (Guiqin Li, German Holguin, Johnny Park), Penn State University (Brian Lehman, Larry Hull), and Washington State University (Vincent Jones) created this database in order to facilitate the research efforts of developing intelligent algorithms for automatically detecting apples with internal feeding worm (IFW) damage.

The database includes two groups of data, 2009 dataset and 2010 dataset. Over 2,700 apple images in 2009 dataset, and over 5,800 apple images in 2010 dataset were collected each year from July to September at the Penn State University Fruit Research and Extension Center. The images have four apple cultivars: Fuji, Golden Delicious, Red Delicious, and York. About a half of the images have apples with IFW damage, and the other half healthy apples. Table 1 shows the number of apples with IFW damage and healthy apple images for each cultivar in 2009. Table 2 shows the same data for 2010.

Each apple in the database is clipped to a 120 by 120 pixels window. Images of apples collected in the same day are assembled into two larger images, one including all the damaged apples, and the other one including all the healthy apples. Image files are named by date (YYYYMMDD), cultivar, and damage condition. For example: 20090701_fuji_damaged.jpg and 20090701_fuji_healthy.jpg

The following image samples are for (a) Fuji, (b) Golden Delicious, and (c) York cultivars. The top row shows the images of damaged apples, and the bottom row the images of healthy apples.

Figure 1 (a) Fuji Variety.

Figure 1 (b) Golden Delicious Variety.

Figure 1 (c) York Variety.

Ground Truth

Ground truth of the IFW damage is provided in two formats: (1) Plain Text and (2) Binary Mask Image.

In the Plain Text format, the file includes the pixel coordinates of the center of IFW damaged regions followed by the approximate radius of the region. The ground truth files can be easily identified by the suffix '_gt'. For example: 20090701_fuji_damaged_gt.txt is the corresponding ground truth text file for the image 20090701_fuji_damaged.jpg.

In the Binary Mask Image format, white pixels (i.e., pixels with 255 pixel intensity value) in the image represent IFW damaged regions while black pixels (i.e., pixels with 0 pixel intensity value) represent healthy regions and background. The Binary Mask Images have the suffix ‘_mask’. For example: 20090701_fuji_damaged_mask.jpg is the corresponding ground truth binary mask image for 20090701_fuji_damaged.jpg

Figure 2. Raw Image 20090701_fuji_damaged.jpg

Figure 3. The Binary Mask Image for 20090701_fuji_damaged.jpg

How to Obtain Access to the Database

To obtain a username and a password for the database, download the license agreement and send a signed copy of the agreement to Somosmita Mitra (somosmita.mitra.1 <at> purdue.edu). Please, use your institutional email and include a short note intruducing your self and the institution you belong to.

Once you have obtained a username and password, click here. If you have any difficulties with accessing the database, please email (rvl-webmaster <at> ecn.purdue.edu).