This paper presents the first outdoor field trials of Deep joint source-channel coding (DeepJSCC) for image transmission over an industrial 5G system, aiming to address the growing demand for efficient and reliable communication in low-SNR environments. DeepJSCC, a deep learning-based joint source-channel coding method, integrates source and channel coding within a unified framework, enabling robust image transmission while mitigating the cliff effect commonly observed in traditional methods. We modify a commercially available 5G base station (gNB) and user equipment (UE) to process DeepJSCC signals, and conduct experiments in both indoor and outdoor environments, including line-of-sight (LoS) and nonline-of-sight (NLoS) conditions. The results demonstrate that DeepJSCC outperforms conventional JPEG2000 and LDPCbased schemes in maintaining image quality in terms of peak signal-to-noise power ratio (PSNR) under extreme transmission conditions. Notably, DeepJSCC achieves superior PSNR stability, particularly in low-SNR regions, and demonstrates the ability to transmit images without corruption, even when the methods employing JPEG, LDPC, and normal QAM modulation fail. Additionally, the system exhibits a lower peak-to-average power ratio (PAPR) compared to typical modulation formats such as QPSK and 16QAM.This preprint is currently under review at IEEE-OJCOMS (IEEE Open Journal of the Communications Society).