The advanced signal detectors cause serious technical problems for detecting signals in orthogonal frequency division multiplexing (OFDM) with index modulation (IM). The existing machine learning (ML)-based signal detection approaches are highly complex. Hence, a deep learning (DL)-based signal detection approaches are proposed in this by improving the detection performance and minimizing the complexity of complex deep neural network (C-DNN) and complex convolutional neural network (C-CNN)-based intelligent signal detection (ISD) technique. The proposed ISD technique is applied for the semi-blind channel estimation (S-BCE) and uses the proposed architectures for the reconstruction of transmitted symbols with the usage of channel state information (CSI). To optimize these network parameters, stochastic gradient descent (SGD) was used to achieve a potential solution for channel estimation (CE) to detect the signals. By simulating the OFDM-IM system for various parameters of bit error rate (BER) and signal-to-noise ratio (SNR). The proposed C-CNN and C-DNN-ISD techniques have achieved 98.99% of accuracy in ISD, 19.39% of BER, and 20.43% of SNR.