Figure 1: An overview of the TCDA dataset. (a) shows the annotation information of the TCDA dataset, (b) shows the viewpoint position of the camera in the classroom during the recording process, and (c) shows examples of different perspectives. The TCDA dataset contains how teachers dress in their daily classrooms，with a total of 11,879 image samples, covering different subjects, classrooms and diverse teaching environments. The details are shown in Figure 1.
As shown in Figure 1(a), we have defined a series of clothing attributes to capture the characteristics of teacher clothing more comprehensively. Positive attribute represents the clothing attribute that the teacher recommends in the daily classroom. Negative attribute represents the clothing attribute that the teacher does not recommend in the daily classroom. Other attribute represents the sample that does not participate in the teacher's dress assessment, and the absent attribute represents the attribute that the character is missing. Finally, we conduct a comprehensive evaluation of the teacher's attire based on the positive attributes and negative attributes to form the final score.
In order to enhance the robustness of the dataset, we try to simulate the shooting angles in the collected classroom teaching videos, i.e., directly behind the classroom, 45° behind the left of the classroom, and 45° behind the right of the classroom,as shown in Figure 1(b). Considering the limitations of the scene, the teacher's body is often partially obscured by the lectern or the teacher's interaction with the students, etc. Therefore, as shown in Figure 1(c), we select the teacher's viewpoint information from four directions, i.e., forward (F), backward (B), left (L), and right (R).
Python >= 3.7
Pytorch >= 1.7.0
1.TCDA dataset If the article is accepted for publication, we will upload the TCDA dataset. Then, prepare the dataset to have following structure: TCDA data/ dataset_all.pkl
2.Training & Evaluation
To train and evaluate the TDAM model on TCDA base on resnet50:
python train.py --cfg ./config/pedes_baseline/TCDA.yaml
If you use the TAQA dataset, please cite this paper:A Teacher Classroom Dress Assessment Method Based on a New Assessment Dataset