Diabetic Foot Ulcers (DFUs) pose a significant healthcare challenge due to their potential to lead to severe complications, including infections and ischemia, often necessitating limb amputation or causing mortality. Accurate diagnosis of ulcer regions, and prompt response can prevent amputations and fatalities through early and reliable DFU detection. However, manual examination of DFUs is labor-intensive and prone to variability. In response to these challenges, we propose DFUC_CAM, a novel method for instance segmentation of DFUs using self-supervised learning (SSL). The proposed approach leverages a multi-stage methodology. Initially, a convolutional neural network (CNN) trained for classification is utilized to generate Class Activation Maps (CAMs), which localize and highlight discriminative regions within DFU images. These CAMs serve as pseudo-labels to train a segmentation network using a SSL framework. For the MICCAI 2024 Diabetic Foot Ulcer Challenge (DFUC 2024), we trained DFUC_CAM using the DFUC 2024 and FUSeg datasets. The FUSeg dataset primarily serves for the quantitative evaluation of our method. During the validation phase of DFUC 2024, our team, mlthings.me, achieved a mean Dice score of 0.342, and in the final testing phase, we improved this to 0.385. These results demonstrate the potential of our approach for enhancing DFU detection and segmentation in clinical settings. The code is available at: https://github.com/RyersonMultimediaLab/DFUC_CAM .