Radio frequency (RF) fingerprinting plays a vital role in enhancing the security of wireless Internet of Things (IoT) networks. Initially, our study offers a comprehensive dataset of RF fingerprinting, collected from 113 LoRa devices operating under indoor communication conditions. The signal was gathered using three varied receivers, specifically RTL-SDR, LimeSDR Mini, and USRP X310. Further, we introduce IQResNet, a pioneering convolutional network that amalgamates IQ correlation features and time features, extracted through residual blocks. IQResNet's efficiency and streamlined design result in an average accuracy of 96.35%. Our research underscores the significance of consistency between training and test data in cross-receiver RF fingerprinting experiments. This is manifest in the enhanced accuracy achieved when data from the same receiver is used. Conversely, test data from divergent receivers causes a marked reduction in accuracy, attributable to the unique RF fingerprint of each receiver.