Final DIRSIG4-only Release and latest DIRSIG4+DIRSIG5 Combo Release

So long DIRSIG4 ...

We are marking the end of the dedicated DIRSIG4-only releases with a final build and packaging of DIRSIG4 using the heritage build hosts that have been producing the DIRSIG4 releases for years. If you have been using DIRSIG4 exclusively, then consider this the final release that has the same end user operating system assumptions. To obtain a copy of this final heritage release, navigate to the "Final DIRSIG4 Stable Release" on myDIRSIG to download DIRSIG 4.7.5.18476.

Hello DIRSIG5!

We are also pleased to announce that the latest DIRSIG4+DIRSIG5 combo release is available. To obtain a copy of this release, navigate to the "Current DIRSIG5 Preview Release" on myDIRSIG to download DIRSIG 5.0.10.3405. DIRSIG5 is mature and used almost exclusively by the student and staff researchers at RIT. However, because DIRSIG5 is not at full feature parity with DIRSIG4 (but exceeds DIRSIG4's capabilities in other areas) DIRSIG5 is packaged with the latest copy of DIRSIG4 as well. How does that overused saying go? "DIRSIG4 is dead, long live DIRSIG4!" But seriously, we strongly encourage all users to start migrating to this newer, faster and more capable version of DIRSIG that has been architected for the next decade.  

DIRSIG5 Advantages

Here is the brief summary of how DIRSIG4 and DIRSIG5 compare:

• DIRSIG5 is multi-threaded and an MPI-enabled build is available for use in parallel computing environments.
• DIRSIG4 was single-threaded and did not not have an MPI-enabled version.
• DIRSIG5 migrated from the classic Whitted-style radiative transfer approach used in DIRSIG4 to the more modern "path tracing" approach.  This allows the model to compute multi-bounce radiances significantly faster.
• For non-trivial scenes, material configurations and collection scenarios, we find that single-threaded DIRSIG5 is anywhere from 20x to 50x faster that DIRSIG4.
• In addition, the quality of the simulation is now controlled by a single mechanism rather than spread across material configurations (radiometry solver parameters) and sensor configurations (sub-pixel sampling, etc.).
• DIRSIG5 supports DIRSIG4-era scene representations through a process called "scene compilation", where the scene is converted from the (sometimes extensive) hierarchy of ASCII/Text geometry and material input files into a single, binary HDF file for run-time loading.
• This scene compilation process only needs to be repeated if the scene is modified.
• By migrating the expensive text-based  file parsing and data verification checks to this rarely performed scene compilation process, the HDF scene representation can be loaded in 10s of seconds compared to 10s of minutes for some complex scenes.
• Neither DIRSIG4 nor DIRSIG5 contain GPU-accelerated code at this time.
• But we have been actively exploring ways to introduce GPU acceleration into DIRSIG5.
• The DIRSIG5 development environment was built with continuous integration from day #1. Continuous integration (CI) means that the code is constantly being compiled and tested by autonomous worker processes to monitor the health of the code.
• Our CI environment provides the developers we nearly instant feedback on the impact of code changes by leveraging a multi-tiered testing system that ranges from quick, radiometric tests that check the low-level code to an extensive set of nightly simulations that provide end-to-end checks on full simulations.
• The CI environment also automates our release builds, which improves consistency in release quality and reliability.
• DIRSIG5's migration to a plugin-centric approach means that the low-level core radiometry is streamlined, optimized and rarely modified. This allows new features to be introduced at a higher level and within the isolation of the plugin framework, which results it improved code stability. Below is a list of plugin-driven features that have been introduced in DIRSIG5:
• BasicPlatform (DIRSIG4 compatibility for platform, motion and task files)
• Improved temporal integration (faster overall, continuous time sampling vs. quantized in DIRSIG4, better direct representation of jitter).
• New point-spread function (PSF) method that directly incorporates the PSF into the object plane sampling contributions (avoids the need to produce oversampled output, convolve and resample).
• New active pixel area sampling, which lets the user provide active area masks for pixels (array constant at this time).
• Support for time-delayed integration (TDI) within the simulation by indicating that a dimension of a detector array is used for TDI collection.
• Additional sensor plugins (SphericalCollector, OrthoImage)
• Additional atmosphere plugins (NewAtmosphere)
• Combination sensor and atmosphere plugins (ChipMaker)

What's missing in DIRSIG5?

At this time, DIRSIG5 still has a set of limitations compared to DIRSIG4. In many cases, these are currently being addressed or will be addressed in the near future. However, some capabilities in DIRSIG4 are not on the current DIRSIG5 roadmap due to a desire to focus on features and performance for the 99% of simulations that users look to the model to perform.

• DIRSIG5 does not have an internal temperature prediction capability at this time.
• During the Winter of 2019/2020, the first of several internal temperature calculation options will be introduced.
• In the meantime, there are a variety of external temperature mechanisms that can be used including the geometry-bound temperatures in GDB files, MuSES geometry+results files, the data-driven temperature solver and temperature maps. Additional internal temperature calculation options
• DIRSIG5 does not support LIDAR/LADAR at this time, but the development effort to enable this modality has just started.
• We expect to see the initial releases with this modality available in the Summer 2020.
• The radiometry solution for this modality will also leverage the "path tracing" solution used for passive illumination, and will eliminate the complexity of correctly configuring the "photon mapping" setup in DIRSIG4.
• DIRSIG5 does not currently support "space situational awareness" (SSA) scenarios at this time.
• The configuration of these scenarios was always "clumsy" in DIRSIG4 and we are currently working on new scene representations that simplify the geometric and atmospheric configurations. We expect to see these scenarios gain support in 2020.
• DIRSIG5 does not support passive polarization simulations.
• This modality introduced a significant amount of code complexity and computational overhead and yet was used by a tiny fraction of the user population. In addition, the remote sensing community in general is devoid of the necessary datasets to support non-trivial polarized scene construction.
• At this time, there are not plans to add support for this modality.
• DIRSIG5 does not currently support real or synthetic aperture radar simulations.
• This was another modality that was not heavily used by the user community.
• At this time, there are not plans to add support for this modality.

With the exception of these current limitations, DIRSIG5 can largely be used as a drop-in replacement for DIRSIG4. A set of plugins (BasicAtmosphere and BasicPlatform) provide compatibility with DIRSIG4-era inputs. At this time, a new graphical user interface (GUI) has not been created to replace the DIRSIG4-era interface. But, the GUI is now DIRSIG5 aware and allows the user to choose which generation of the model to use to execute a simulation.

Migrating from DIRSIG4 to DIRSIG5

If DIRSIG5, meets all your needs now, then we encourage you to try out the new software. The DIRSIG4+DIRSIG5 combo releases can be found on myDIRSIG. To learn more about migrating from DIRSIG4 to DIRIG5, we strongly encourage you to read the DIRSIG5 Migration Manual.

One of the easiest ways to determine the level of support or compatibility of a given feature in DIRSIG is to look at the list of demos. Each demo is now badged with a set of one or more icons that indicate the compatibility of the demo with either DIRSIG4 or DIRSIG5.