In many different industrial, residential, and commercial situations, gas leakage poses a serious hazard. Its discovery is crucial since it might have severe effects like explosions and fires. For the protection of persons and property, as well as to avert catastrophic tragedies, accurate and prompt gas leak detection is essential. Convolutional Neural Networks (CNNs), in particular, have demonstrated encouraging results in the detection of gas leaks in recent years. Here, a CNNbased method is provided for detecting gas leaks from image data. The suggested method employs a Softmax classifier for gas classification after extracting features from the image dataset using a combination of convolution, pooling, and fully connected layers. The usefulness of the suggested approach in accurately detecting gas leakage is shown by the experimental findings and the proposed approach is tested on a real-world gas leakage dataset. It can be added to gas detection systems to improve their functionality, lowering the likelihood of gas-related mishaps. The findings support current work to create accurate and effective machine learning-based gas leak detection systems.