This COCO-Bridge_readme.txt file was generated on [2019-04-28] by [Eric Bianchi] ------------------- GENERAL INFORMATION ------------------- 1. Title of Dataset: ***COCO-Bridge, structural details for bridge inspection 2. The author, a Virginia Tech graduate and PhD student, ERIC BIANCHI in the SEM (structural engineering and materials) program in the college of civil engineering. Principal Investigator Contact Information Name: Matthew Hebdon Institution: Virginia Tech, SEM, college of civil engineering Address: 105B Patton Hall, 750 Drillfield Dr. Blacksburg, VA 24061 Email: mhebdon@vt.edu Associate or Co-investigator Contact Information Name: Pretap Tokekar Institution: Virginia Tech, ECE, college of electrical and computer engineering Address: 453 Whittemore, Virginia Tech, Blacskburg, VA 24061 Email: tokekar@vt.edu Alternate Contact Information Name: Eric Bianchi Institution: Virginia Tech Address: 185 Sequoia Cir, Christiansburg VA, 24073 Email: beric7@vt.edu 3. Date of data collection: ***from (2018-08) to (2019-03) 4. Geographic location of data collection (where was data collected?): ***Norfolk Southern Railway yard in Roanoke Virginia and from engineering firms (Clark Nexsen, VDOT) 5. Information about funding sources that supported the collection of the data: ***National Science Foundation (NSF) -------------------------- SHARING/ACCESS INFORMATION -------------------------- 1. Licenses/restrictions placed on the data: ***Cite the Dataset as outlined in #6 (Recommended citation for the data) 2. Links to publications that cite or use the data: ***(Currently, no other publications use this data) 3. Links to other publicly accessible locations of the data: ***(http://COCO-bridge.com, https://github.com/beric7/COCO-Bridge) 4. Links/relationships to ancillary data sets: ***There are no ancillary data sets included at the moment 5. Was data derived from another source? If yes, list source(s): ***NO. 6. Recommended citation for the data: ***(Bianchi, E. (2018).COCO-Bridge [Dataset]. University Libraries, Virginia Tech.) --------------------- DATA & FILE OVERVIEW --------------------- ***This dataset was the control dataset in the experiment. 1. Folder List A. Folder Name: Annotations - contains all bounding box annotations for evaluation and training photos B. Folder Name: Evaluation_Photos - contains all the evaluation photos, re-scaled. C. Folder Name: Training_Photos - contains all the training photos, re-scaled. 2. Relationship between files: ***The relationship between all files are outlined on the COCO-Bridge github and website 3. Additional related data collected that was not included in the current data package: ***N/A 4. Are there multiple versions of the dataset? yes/no ***(NOT YET) -------------------------- METHODOLOGICAL INFORMATION -------------------------- 1. Description of methods used for collection/generation of data: ***Gathered images from engineering firms that conduct bridge inspections and self-taken. Followed annotation documentation that was created as a result of a previous effort. These annotation guidelines should be carefully followed to to match the standards of the dataset. Data was annotated using bounding boxes, manually by the researcher. This is important since the structural engineer will provide the most accurate and precise bounding boxes ground truths and may be less likely to miss a structural detail. Further research will work on the automization or un-supervised learning. Also, see website (http://COCO-bridge.com) 2. Methods for processing the data: ***Tensorflow API, use a GPU and parallel computing 3. Instrument or software-specific information needed to interpret the data: ***LableImg python software to read XML files. Must convert CSV files to XML file format to be read in LableImg, but not needed for using files in tensorFlow. TensorFlow 1.8.0 version was used. V-100 nodes were used to train the model. 4. Standards and calibration information, if appropriate: ***Images have been reduced to a size which can be handled by economically by GPUs 5. Environmental/experimental conditions: ***N/A 6. Describe any quality-assurance procedures performed on the data: ***Data is annotated in accordance to the annotation guidelines provided in on the dataset website. 7. People involved with sample collection, processing, analysis and/or submission: ***Eric Bianchi, Dr. M Hebdon and the Virginia Tech Advanced Research Computing (ARC) group at Virginia Tech