Coding genes constitute only a small portion of the human genome. Recently, circular RNAs (circRNAs), commonly thought to be noncoding, have been demonstrated to be translated into novel proteins, which might mediate diverse biological processes. However, it remains challenging to identify translatable circRNAs and circRNA-coded proteins. Here, we introduce CICADA, an algorithm capable of evaluating the coding ability of circRNAs and inferring circRNA-coded products. We find that CICADA's performance is highly efficient and cost-effective compared to existing algorithms. Moreover, we used CICADA to successfully identify functional translatable circRNAs in esophageal squamous cell carcinoma and construct circRNA translation profiles across various cancers. CICADA provides an efficient tool to explore the hidden proteome encoded by translatable circRNAs.