H.R. 3699, the Research Works Act, was introduced Dec. 23, 2011, by Rep. Darrell Issa (R-Calif.), chairman of the Committee on Oversight and Government Reform, and committee member Rep. Carolyn Maloney (D-NY). According to the Association of American Publishers (AAP) website, “The legislation is aimed at preventing regulatory interference with private-sector research publishers in the production, peer review and publication of scientific, medical, technical, humanities, legal and scholarly journal articles.” Put another way, it is designed to thwart activities such as the National Institutes of Health (NIH) Public Access Policy, which requires scientists to submit final peer-reviewed journal manuscripts that arise from NIH funds to the digital archive PubMed Central upon acceptance for publication.

In the harsh words of The Guardian, “This is the moment academic publishers gave up all pretence of being on the side of scientists.” I, for myself, never believed their good intentions.

I you are a resident of USA, please take the time to sign the petition against RWS.

I remember when shareware was a novel concept, and I used to download it by the dozens : nagging screens were part of the deal, something to be expected until you paid for a registered version. Since when those “features” have become cricket for bought wares ?

After wondering if I should open a support ticket, I’ve just reported the pestering behavior as a bug. Lets see if the development team will agree.

Balancing a faculty career, with research, teaching and administrative obligations is more challenging than people outside academia usually realize. For the last 5 years, I was exclusively focused on research, so I am rediscovering the thrill of being in a classroom. I am also discovering the painstaking work needed to sustain academic institutions, for, if their horizontal, democratic nature warrants their members many freedoms, they require in return much debate, discussion and politics.

Nevertheless, I am loving every minute of my new duties. I know that the passing years take their toll, but for the moment, at least, I am in my element.

This has lasted for the last few months, so it’s not just a rotten update: it’s a whole series of rotten updates. It has become so serious that before I found the problem I considered abandoning Firefox altogether. Nowadays I just keep Flash disabled, except when absolutely needed.

Interestingly, living without Flash has been much less nightmarish than I anticipated. Half of the time, even embedded YouTube videos will play (once you sign up for the HTML5 beta). Often, the only thing lost are the more aggressive forms of advertisement. (To which I say good riddance, of course.)

Like, when I visit a (not that popular) website, type another URL a few minutes later and get an ad from the previously visited website ? Coincidence ?

Apparently the detrimental effect this has on consumers is old news. I can attest that I was not positively impressed.

Bossa: Extended BoW Formalism for Image Classification
In image classification, the most powerful statistical learning approaches are based on the Bag-of-Words paradigm. In this article, we propose an extension of this formalism. Considering the Bag-of-features, dictionary coding and pooling steps, we propose to focus on the pooling step. Instead of using the classical sum or max pooling strategies, we introduced a density function-based pooling strategy. This flexible formalism allows us to better represent the links between dictionary codewords and local descriptors in the resulting image signature. We evaluate our approach in two very challenging tasks of video and image classification, involving very high level semantic categories with large and nuanced visual diversity.

I’ve also had two papers accepted on our counterpart national conference, SIBGRAPI. The first is the work of the Ph.D. student Ana Lopes and her R.A. Elerson Santos (supervised by Prof. Arnaldo Araújo and co-supervised by Prof. Jussara Almeida; I give her some technical and nontechnical support every now and then). It concerns the use of transfer learning of concepts from (static) image datasets to video datasets in order to recognize human actions. We show that learning the concepts present on the Caltech256 dataset allow a classifier to obtain improved results on the challenging “in the wild” human action Hollywood2 dataset.

Transfer Learning for Human Action Recognition
To manually collect action samples from realistic videos is a time-consuming and error-prone task. This is a serious bottleneck to research related to video understanding, since the large intra-class variations of such videos demand training sets large enough to properly encompass those variations. Most authors dealing with this issue rely on (semi-) automated procedures to collect additional, generally noisy, examples. In this paper, we exploit a different approach, based on a Transfer Learning (TL) technique, to address the target task of action recognition. More specifically, we propose a framework that transfers the knowledge about concepts from a previously labeled still image database to the target action video database. It is assumed that, once identified in the target action database, these concepts provide some contextual clues to the action classifier. Our experiments with Caltech256 and Hollywood2 databases indicate: a) the feasibility of successfully using transfer learningtechniques to detect concepts and, b) that it is indeed possible to enhance action recognition with the transferred knowledge of even a few concepts. In our case, only four concepts were enough to obtain statistically significant improvements for most actions.

The second is the work of my Ph.D. student Marcelo Coelho and his R.A. Cássio dos Santos Jr. (again his main supervisor is Prof. Arnaldo de Araújo). It concerns the clean-up of noisy SIFT features of street-view images (urban façades). We have found out that subspace clustering, a non-supervised technique, is able to isolate clusters of useful and non-useful SIFT features for the task of retrieving a target image. The challenge is identifying a priori which cluster is the relevant one. This work compares and contrasts two subspace clustering techniques: FINDIT (based on dimension voting) and MSSC (based on a fuzzy mean-shift).

Subspace Clustering for Information Retrieval in Urban Scene Databases
We present a comprehensive study of two important subspace clustering algorithms and their contribution to enhance results from the difficult task of matching images taken of the same object using different devices at different conditions. Our experiments were done on two distinct databases containing urban scenes which were tested using state-of-the-art matching algorithms. After initial evaluation of both datasets by that procedure, clustering algorithms were applied to them. An exhaustive comparison was performed in every cluster found and a significant amelioration in the results was obtained.

I’ll put a link to the preprints as soon as I have they become available.

An interesting difficulty I have faced for reporting on this work was showing representative images, without hurting the sensibilities of reviewers and readers. So far, my (admittedly coward) choice has been taking the tamest images that are still representative of the phenomena I want to illustrate. For example: to illustrate that the dataset is ethnically diverse, I would chose frames where only the faces of the actors are shown; to illustrate that the dataset contains gay porn as well as straight porn, I would show a frame with the actors kissing instead of having sex; etc.

But recently, I had a tough choice to make. A student was to submit his Master disstertation to the viva-voce committee, and, as it usually happens in Brazil, he has sent me a draft for corrections and suggestions. His “Results” chapter contained, among cold graphs and tables, several very explicit images, illustrating in detail the cases of success and failure of our algorithm. The only thing is: all images contained censor bars.

I returned the draft with several corrections, among which, a note begging him to remove the bars:

“Don’t censor the images — it’s extremely distasteful: this is a scientific work for an adult audience. Either remove the images entirely (if they are not needed), either keep them uncensored (don’t mess up with the data !). In the worst case, put them in an Annex or in a separate suplement.”

In the end, he’s decided to keep the images uncensored, which I feel was the right scientific decision.

Nevertheless, everytime I open his “Experimental Results” chapter I cringe a little bit. Againg admittedly cowardly I am looking forward for the defense, when I’ll be able to share the responsibility for the final decision — keeping or taking away the images from the definitive version — with the rest of the committee.

* * *

Taking a (superficial) look in the literature, I noticed that many authors (including myself) practice a form of “partial self-censorship”: choosing “tame” images, making them tiny in the page, or using washed out grayscale reproductions — a compromise between scientific truth and respect to the taboo ? Or just plain cowardice ? Most authors simply don’t include images, and a few choose to employ the censor bars. The full-fledged honesty of my student is rare.

The censor bars, IMHO, are the worst choice — at once hypocritical and unscientific. Hypocritical, because the reader can perfectly imagine what is behind them, so any of the “dirtiness” from which they would be supposedly “protecting” the reader is still being created in his or her mind. The effect is exactly the same as when using euphemisms like “f-word”: the  correct word is still created in the listener mind. Unscientific, because they count on the reader imagination (with its distortions, imprecisions, and, often, amplifications) instead of depicting precisely the phenomena at study.

Interestingly, in one paper, the authors censor the faces of the actors (by pixelization). This is an interesting choice and raises a question I have not considered: since we collect our dataset from pornography sharing social networks, we cannot assume that everyone in the video is a professional actor. I hope that none of our examples have Computer Vision scientists unaware that their amateur videos have escaped to the net !

* * *

In the end of the day, this is 2011 — 64 years since the first Kinsey report ! Shouldn’t science have got some guts by now ?

The system is so unreliable that my only option is to reformat the machine and restart from scratch (basically what I had to do when I used XP — but XP usually lasted a good year and a half between reformats).

With that and all the software incompatibility problems, I am seriously considering switching to Windows 7 + Cygwin and thrashing this whole Mac OS experiment in the huge pile of “tried: didn’t work”.

More and more, for better and for worse, MacOS is the new Windows.