Katherine Gemperline
Week 10 Posted on July 23, 2019 On Monday, we ran the survey on Amazon Mechanical Turk. However, I was not aware there was a way to say we only want participants with an approval rating over a certain percentage. Because of this, we got a lot of bad answers that we then had to throw out. On Tuesday, I started learning RStudio. The graduate student, Alex, who was working with me throughout all of this, helped code as I had never coded in R before. Since my last day is Thursday, I learned the basics of R, but she did a lot of the coding while I asked questions. We also debated running the experiment again because it looked like we had to throw out close to 30% of the participants’ answers. We decided to not run the survey again, and make due with the results we got and just remove anyone with outlying responses. On Wednesday, I made minor edits to my poster and then printed it. I also started finalizing my Research Paper. I corrected the edits that my mentor found, and then I added on the Results and Future Works sections. On Thursday, I presented my poster at the Symposium for Undergraduate Research Students at Clemson University. After, I came back to the lab to continue working on the paper. Thursday was my last day of the DREU at Clemson University. This was a great experience for me, and I am so grateful to everyone involved for making me feel so welcome and for teaching me so much. Week 9 Posted on July 17, 2019 Monday and Tuesday, I made minor edits to the survey. Edits included wording and consent forms, etc. I started creating a poster for the Poster Symposium for REU students at Clemson which included taking screen shots of all the videos. Wednesday, Thursday, and Friday, made more edits to the Survey. I had to modify all the videos I created in Adobe Premiere Pro so that the logo was edited out. The logo was from the script that we used to export the videos in Unity. On Friday, we had hoped to upload the survey to Amazon Mechanical Turk but the person in charge of funding was out of the office. Week 8 Posted on July 8, 2019 Monday, I exported the videos from Adobe Premiere Pro. Each of the finished videos consisted of four of the videos exported from Unity. If I uploaded the videos individually to Qualtrics, where I’m creating the survey, they would not play at the same time. Since that’s pretty important to the survey, I merged the videos into one using Premiere Pro. Tuesday, I finished exporting what was left of the videos in Premiere Pro. I started creating practice surveys in Qualtrics to start narrowing down what distance metrics to use and also what videos we should be comparing. Wednesday through Friday, I continued creating videos in Premiere Pro and uploading them into a Qualtrics survey. At this point, we also decided to use four distance metrics: 2.5, 5, 7.5, and 10. Each distance metric value version was compared to original twice: once with the original on the left and a second time with the original version on the right. Previously, we had decided to compare the original to 2.5, original to 10, 2.5 to 10, and vice versa. This new method allowed for more distance metric values to be tested. Week 7 Posted on July 1, 2019 I started this week by checking the work I did on Friday. There might have been inconsistencies in the Maya Files I used to create the offset animations. I got that fixed and started double checking I’m not adding offsets to versions of the original animation that are already off. I will be showing one animation from two angles: a frontal view and a view from the right, from the viewer’s perspective. I also decided that the distance metrics I use won’t be over 15, because 15 already causes a lot of differences in some gestures– not so much in others, though, in my opinion. But I haven’t decided exactly which ones to use: 5 and 10, 5 and 15, 2.5 and 10, or 2.5 and 15. I still plan on only showing two, though. Then I started creating the survey using Qualtrics, and I spent some time learning how to use it. The main problem I had there was that the videos I input into the questions didn’t show up in the published version of the survey. I tried uploading the videos to my YouTube account and then just embedding them in the survey from Youtube. That didn’t work. The issue was that I didn’t have a licensed version of Qualtrics associated with my account. While I was waiting to ask fellow grad students that weren’t in the office on Monday and Tuesday to ask them questions about Qualtrics, I started making the rest of the videos. I wanted to check that the camera angles, lighting, and motions themselves were good enough for the survey. Week 6 Posted on June 24, 2019 This week I finished the Python script for mirroring the left hand onto the right hand. I spent the rest of Monday, trying to find a clip of a hand grasping without the arms moving too much. That has been the hardest clip to find. In the end, I went without that motion. I revisited the animations of “small” and “okay.” I disregarded both previously because the crease at the wrist didn’t fold correctly with the old models. Since I changed the skin weights in Maya since then, they were fine to use. On Tuesday, I spent the day fixing the lighting in the animations. Then I fixed a pinky joint that was at a weird angle. I searched for more clips to use so that when I finally design the experiment, I have more options to choose from. On Wednesday, I changed the mesh for the model back to the original model. So then, I started thinking about the set up of the experiment and also started creating the videos of all the edited versions of the motions. Thursday, I started by finding the degree of changes I would make in the animations. I narrowed the base motions down to four: attention, idle, small, and shrug. Next I started figuring out which values for the distance metric to use: 5, 10, 15, 20, 25, or 30. Since, I’m only planning on having two, I’m thinking 5 and 10. They are closest to appearing normal. We are planning on using two at this point in time. We would be comparing all distance metric values to all the distance metric values and therefore, there would be too many comparisons for a single participant to make, even if they only answer comparisons about one base motion. Week 5 Posted on June 19, 2019 This week started with me finding better motions again where the arm doesn’t move as much. I have a few ideas on which ones to use, but the issue lies in the motions with the least amount of arm movement have a different rig in Maya than the previous motions I used. I spent Tuesday figuring out different ways to bind the model to the rig; the main issue that arose was that there were two joints in the wrist in the skeleton in Unity, but the animation only has one wrist joint. This caused the hands in the animation to be static. However, at the end of the day, I got the animation working. The new problem is the skin on the animation is weighted for two joints. On Wednesday and Thursday, I spent the entire day, messing with the weights in the hands so that the hands appeared natural in regards to anatomy. In the afternoon on Thursday and then on Friday, I spent time creating a script that mirrors the animation on the left hand to the right hand. I was able to get everything but the fingers mirrored. Week 4 Posted on June 10, 2019 This week, I recorded some initial videos and kept track of distance metrics as I went through the clips. At this point in time, the range of distance metrics I’m looking at is .5 to 50. On Tuesday of this week, I created new scripts in Python to be used on Maya animations which will cut down on the time it takes to modify and record the videos. The scripts were the same in nature from what I was doing before in Unity; I am still using jitter, added motion, offset in the hand, and offset in the fingers. I am still recording in Unity, but adding the modifications, or errors, in Maya. The scripts in Python export the angles to .csv files, and when I find the correct distance metric, I export that new animation to Unity and record it. Since the animations of the hands were taken from animations that were actually an entire person acting out charades for a movie, The Pianist, I cut off the animations at the hands. They looked weird, though, so I removed the animations for the hands, and just left the fingers and thumb. That too looked weird, so I added the wrist and hand motion back into the animation, and while it may not look as weird, the hand moves in and out of frame too much. To counteract the hands moving too much, we had the camera follow the hand. This created a different problem where the hand was stagnant on the screen, but the forearm was swinging all around the perimeter of the screen. Week 3 Posted on June 4, 2019 I started week 3 by finishing my code to calculate the distance metric between two animations. The Euler Angles were outputted to excel files and are then able to be read by the code. The code systematically checks all lines of the excel file. Each line represents a single frame. The columns represent the x, y, z rotation of each joint. So each joint is represented in three columns. The code checks similar frames- frame 1 from both animations, then frame 2 from both animations, then frame 3 from both animations, and so on. When the same files is inputted twice, i.e. comparing the file to itself, the distance metric is then 0. Once the Distance Metric code was finished. I started writing the research paper that will be located under the Research Paper tab on this website. Dr. Joerg encouraged me to start the paper earlier, specifically the introduction and methods sections, as it spaces out the workload. To write the paper, I used LaTex, a software completely new to me. I also attended two workshops. One on the publication process for research papers and the other on writing personal statements for professional development. The person leading the Personal Statement Workshop explained what to do and what not to do, but also explained the admission process on the school’s side. He also explained the different paths that someone could go through from undergrad to graduate school. Week 2 Posted on May 28, 2019 This week, I started by asking more questions about Unity and Maya, mostly in regards to the animations that were given to me. The animations were used in some of the grad students’ past research projects, and I’m going to be modifying them to fit my research project since I don’t have time to animate my own. The Distance Method I will be using is based on the rotations of the joints of the hands using Euler Angles to calculate the difference between two animations. I will be comparing two animations: an original animation and a modified version of that animation, such as adding an offset in all angles in the hand, an offset of one angle, adding a motion to the fingers, and creating a jitter. Week 1 Posted on May 22, 2019 I started off the DREU getting set up at a desk and meeting everyone. I read articles the graduate students I’m working with had written and downloaded programs, such as Unity. I talked with my mentor about my own personal project along with another project I would assist on. The other project uses a VR escape room that will be used to test different conditions that haven’t actually been determined yet. The escape room was still being debugged. The personal project I will be creating is using distance metrics and perception to show that the common distance metric based on Root Mean Square Error isn’t accurate when it comes to a viewer’s perception. This creates a problem when searching large databases of motions which is common in developing video games where motions are often reused. To get started learning Unity, I created a version of Brick Breaker using an online tutorial. I also went through Citi Training in preparation for the research aspect of my project. This was a good experience for me, since I learned a lot of the process of IRB and applying for research that involves people. At the end of the week, I officially started my own project using Unity and Maya.