Dataiku enables users to write Keras models in the Deep Learning lab.
You can code any CNN architecture that is compatible with the keras and tensorflow versions of your code environment (e.g. keras==2.3.0 and tensorflow==1.15.0 for dss 8.0.4)
You can have a look a the course on dataiku's academy plateform to see how to build your own network.
If you are looking for pre-trained models:
- and your are doing Image classification you can have a look at the Image classification plugin documentation that will download and finetune pre-trained : ResNet, Xception, Inception V3 or VGG16 models.
- and your are doing Object detection you can have a look at the Object detection plugin which will download and finetune a retinanet model.