CStudentAct Dataset

Although several image and video datasets have been collected for student activity recognition, they are mainly annotated at the frame level or sequence level, which can be used for object detection-based and video classification-based approaches. Therefore, to promote research in continuous student activity recognition, we have prepared a new dataset named CStudentAct. The CStudentAct Dataset is an extension of the StudentAct Dataset. It provides both temporal and spatial information for each activity. To build CStudentAct, we have annotated sequences for all bounding boxes belonging to the same action instance by linking them frame-by-frame over time. Figure 1 shows an example of labeling the raising_hand activity.

Example of labeling the raising hand activity sequence in CStudentAct dataset

Figure 1: Example of labeling the raising hand activity sequence in CStudentAct dataset

Details

Each labeled data sample includes the following information:

Activity type
Action instances
Bounding boxes of the activity in each frame of interest

In this dataset, action instances are defined as the names of dynamic activity sequences of students in the classroom. This means that the same person can have multiple action instances, each representing a different sequence of activities. Figure 2 illustrates several labeled activity sequences.

Data acquisition system

Figure 2: Some of examples of activity sequences in our CStudentAct dataset: (a) raising hand, (b) standing, (c) sleeping, (d) using phone

After reviewing all activity sequences, the results are stored in a text file containing: frame ID, action instance ID, x, y, w, h, as shown in Figure 3. This serves as the basis for later evaluations. Using action instances helps to describe student behavior in the classroom in detail. Compared to using only image-level labels, action instances provide more information about how, where, and when activities are performed.

file structure stores labeled results of the Sleeping

Table 1 shows the number of images, the number of action instances, and the number of labeled bounding boxes for different activity types in our CStudentAct dataset:

Activity	Number of images	Number of act_ins	Number of bbox
Raising_hand	296	39	3991
Using_phone	3125	26	11998
Sleeping	3313	31	10162
Standing	4623	46	6626

The CStudentAct dataset is meant to aid research efforts in the general area of developing, testing and evaluating algorithms for human activity recognition. The Hanoi University of Science and Technology (HUST) has copyright in the collection of activity video and associated data and serves as a distributor of the CStudentAct dataset.
Release of the Dataset To advance the state-of-the-art in activity recognition, this dataset could be downloaded. The requestor must sign in the commitment and send it to the dataset administrator (lan.lethi1@hust.edu.vn) by email. In addition to other possible remedies, failure to observe these restrictions may result in access being denied for the dataset.

The researcher(s) agrees to the following restrictions on the CStudentAct dataset:

Redistribution: Without prior written approval from the dataset administrator, the CStudentAct dataset, in whole or in part, will not be further distributed, published, copied, or disseminated in any way or form whatsoever, whether for profit or not. This includes further distributing, copying or disseminating to a different facility or organizational unit in the requesting university, organization, or company.
Modification and Commercial Use: The CStudentAct dataset, in whole or in part, may not be modified or used for commercial purposes.
Publication Requirements: In no case should the still frames or video be used in any way that could cause the original subject embarrassment or mental anguish.