<p>Efficient image compression with reliable transmission and effective rate-control is critical for implementation in a resource constrained wireless environments like Internet of Things (IoT) applications. The regular compression techniques are often suffered with reduced compression ratio that leads to utilization of excessive bandwidth in the network environment. Moreover, the traditional transmission protocols lack in terms of adaptive capabilities and that results in increased packet loss leads to poor packed delivery ratio. In order to address this issue an AI-driven Adaptive Compression and Transmission (AIACT) framework is designed by integrating deep learning-based image compression algorithms with an intelligent transmission control mechanism. The work includes light weight Convolutional Neural Networks (CNNs) for extracting the image features and a reinforcement learning based technique is used for adaptive transmission decisions. The proposed methodology is designed to ensure an optimal bandwidth utilization, effective rate-control and minimal distortion. The performance of the proposed AIACT framework is evaluated with Kodak and DIV2K datasets to prove its effectiveness on image compression and transmission over the traditional methods. For the experimental work, the compressed image data are converted into data packets with respect to the payload of IEEE 802.11 protocol and fed into the network simulator model design in MATLAB simulation. The experimental analysis indicates a better SSIM of 0.869 at compression ratio of 6.1 in Kodak dataset with packet delivery ratio of 92.3% and the same has given 0.903 SSIM at compression ratio of 6.8 in the DIV2K dataset with a packet delivery ratio of 93.3%.</p>