The paper presents a planar diffractive neural network (pla-NN) designed for direct electromagnetic (EM) signal processing in the microwave frequency range. The pla-NN is characterized by its highly integrated and conformal architecture, which overcomes the limitations of traditional diffractive neural networks, such as misalignment and large size. The network is fabricated using printed circuit technology, ensuring high fabrication accuracy and flexibility in design. The pla-NN consists of propagation layers and modulation layers, with couplers and phase shifters, respectively, enabling flexible extension and stacking configurations. The network's performance is validated on the fashion-MNIST dataset through numerical simulations and experimentally verified using a system for object recognition in EM space. The pla-NN demonstrates high classification accuracy and robustness to misalignment, making it suitable for applications in high-performance computing, wireless sensing, and flexible wearable electronics. The paper also discusses the processing speed, scalability, reconfiguration ability, nonlinear function, and on-site training capabilities of the pla-NN, highlighting its potential for advanced signal processing techniques in the microwave frequency range.The paper presents a planar diffractive neural network (pla-NN) designed for direct electromagnetic (EM) signal processing in the microwave frequency range. The pla-NN is characterized by its highly integrated and conformal architecture, which overcomes the limitations of traditional diffractive neural networks, such as misalignment and large size. The network is fabricated using printed circuit technology, ensuring high fabrication accuracy and flexibility in design. The pla-NN consists of propagation layers and modulation layers, with couplers and phase shifters, respectively, enabling flexible extension and stacking configurations. The network's performance is validated on the fashion-MNIST dataset through numerical simulations and experimentally verified using a system for object recognition in EM space. The pla-NN demonstrates high classification accuracy and robustness to misalignment, making it suitable for applications in high-performance computing, wireless sensing, and flexible wearable electronics. The paper also discusses the processing speed, scalability, reconfiguration ability, nonlinear function, and on-site training capabilities of the pla-NN, highlighting its potential for advanced signal processing techniques in the microwave frequency range.