Packet collisions due to overlapping Basic Service Set (BSS) interference can result in decreased network efficiency. To overcome this problem, 802.11 stations need to confirm the channel availability through the Clear Channel Assessment (CCA) mechanism. Besides energy detection, which has low sensitivity, packet detection in Wi-Fi also utilizes a more sensitive preamble correlation method to check for a valid data symbol within 4µs. Hence, a successful CCA is only possible if the received signal exceeds a certain power level or the contending station could receive the signal preamble. In this paper, we demonstrate how a deep learning-based method like Convolutional Neural Network (CNN) can be used to supplement Wi-Fi’s CCA capability to increase its sensitivity without requiring the reception of the signal preamble. We also elaborate on the role of filter size and max-pooling in signal processing points of view. Our simulations show that the proposed CNN model can classify different packet formats and their modulation types, even when different symbol timing offsets occur. Based on the ns-3 simulation, we demonstrate that the proposed approach leads to better channel utilization and enhanced throughput in a multiple BSS network by prioritizing high throughput transmissions.