In classical machine learning algorithms, diagnosing patients with Coronavirus Disease-19 (Covid-19) from Chest X-ray images is a major problem. This research proposes to use pneumonia photos to increase the accuracy of various patient categories using Improved Inception ResNet-v2. An Inception-ResNet-v2-based multiscale channel attention module is shown to provide network recognition and target detection in the face of drastic scale shifts. At the beginning of the model, the effective receptive field and the convolution kernel of the network stem layer are both larger. The activation function is made smaller and the SiLU activation function is utilized rather than the ReLU activation function in order to prevent the model from being overfit. In order to solve the problem of the Chest X-ray dataset having less data, the input image