Figure 4. Statistical assessment method. The statistical test measures the likelihood that two samples are similar. Measure this by calculating their distance concerning their histograms. A high value in the KS test shows a high probability of two samples being drawn from different populations or being statistically different. Experiments The experiments were performed on a server with Intel (R) Xeon (R) Silver 4110 CPU @ 2.10 GHz, 132 GB random access memory (RAM), and running on Ubuntu 16.04.6 LST 64-bit operating system. The proposed fisherman action recognization model was implemented using the tensor-flow framework. To detect fishermen, we developed a customized YOLOv7 model. We retrained the YOLOv7 model using transfer learning with a dataset of 603 training images, 172 validation images, and 87 testing images. The dataset was created from CCTV video footage of vessel CT-1267 during its voyage period from 1 November 2021 to 21 March 2022. Twenty-eight fishermen were working on the CT-1267 vessel, with 12 CCTVs installed. Only four of the CCTVs captured footage of the main working area. For the experimentation, we focused on CCTV3 and CCTV4. CCTV3 captured the working area where fishermen collected and cleaned fish from the net. Figure 5 illustrates the working area of CCTV3 and the detection of fishermen engaged in the fish collection work. CCTV4 captures the work area where the fisherman performs the baiting work. Figure 6 depicts the working area of CCTV4 and fisherman detection September 2023 of the baiting work area. The statistical assessment method is utilized to identify instances of labor exploitation between 2 June 2022 and 2 September 2022. The acceptable threshold for the p-value for this experiment is 0.05. The null hypothesis is " there is no concern of over-exploited laboring. " The Figure 5. Working area of CCTV3 of the fish collection work area. Figure 6. Working area of CCTV4 of the baiting work area. 93