Py4J is developed by Barthélémy Dagenais. The goal is to enable developers to program in Python and benefit from Python libraries such as lxml while being able to reuse Java libraries and frameworks such as Eclipse. You can see Py4J as an hybrid between a glorified Remote Procedure Call and using the Java Virtual Machine to run a Python program.
As opposed to Jython, Py4J does not execute the Python code in the JVM so it does not need to reimplement the Python language and developers can use any libraries supported by their Python interpreter such as libraries written in Cython. Py4J is more focused than JPype. For example, it does not attempt to link Python threads to Java threads. JPype also uses JNI to communicate with the JVM while Py4J uses plain old sockets, which are more portable in practice.
In terms of performance, Py4J has a bigger overhead than both of the previous solutions because it relies on sockets, but if performance is critical to your application, accessing Java objects from Python programs might not be the best idea :-) We plan to evaluate the performance overhead of Py4J once we have a stable release.